* master:
  component/ruby: download source on https
  component/helloweb: version up (to fix gem building with ruby 2.6)

* master:
  software/rapid-cdn: make re6st-verification-url optional
  software/ors-amarisoft: enb.jinja2.cfg: nr: Move tunable parameters from nr_cell_default to nr_cell_list
  software/ors-amarisoft: enb.jinja2.cfg: lte: Move tunable parameters from cell_default to cell_list
  software/ors-amarisoft: slapos-render-config: Replace *arfcn numbers with valid ones
  software/ors-amarisoft: ru/*: Wrap with `if not testing` things that require real environment
  software/ors-amarisoft: ru/*: Pull testing initialization and default into one place
  software/ors-amarisoft: Merge ue-lte.jinja2.cfg and ue-nr.jinja2.cfg into ue.jinja2.cfg
  software/ors-amarisoft: slapos-render-config: Also render ue-*.cfg
  software/ors-amarisoft: slapos-render-config: Emit rendered files into config/out/
  software/jstestnode: serve test files on https
  component/ruby: remove version 2.3
  component/nodejs: remove version 8.12.0
  software/gitlab: upgrade to version 12.10.14
  component/golang: golang 1.13 is needed for gitlab 12
  component/coreutils: add version 9.1 which is needed for gitlab-backup
  component/ruby: add version 2.6 needed by gitlab 12
  Use most recent osi-coupler version.

software/ors-amarisoft: enb.jinja2.cfg: nr: Move tunable parameters from nr_cell_default to nr_cell_list

This patch is similar and has the same rationale as the previous patch
(software/ors-amarisoft: enb.jinja2.cfg: lte: Move tunable parameters
from cell_default to cell_list). Please see its description for context
and details.

Like with previous patch I've made sure via json_util that effective gnb
configuration stays exactly the same. Also, since for NR case we move
more code, and also use enb merging feature(*) I tested that the
effective gnb configuration stays exactly the same with both

  * tdd_config = 3  (5ms 8UL 1DL 2/10 (maximum uplink)),	and
  * tdd_config = 1  (5ms 2UL 7DL 4/6 (default))

In all the cases effective gnb configuration stayed exactly the same
before and after this patch which proves that effective configuration
for single cell case remains unchanged.

(*) enb implements merging process for cell/cell_default and
nr_cell/nr_cell_default as demonstrated below:

  cell={a: 10, d:15}  cell_default={a:1, b:2, c:3}       ->  cell={a:10, b:2, c: 3, d: 15}
  cell={x: [{}, 2]}   cell_default={x: [{a:1}]}          ->  cell={x: [{a:1}, 2]}
  cell={x: [{c:11}]}  cell_default={x: [{a:1}, {b:2}]    ->  cell={x: [{a:1, c:11}, {b:2}]}

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1515

Currently we set some cell tunable parameters on the cell object itself
(e.g. dl_earfcn), while other cell tunable parameters on the
cell_default object, e.g. cell bandwidth. This works ok when there is
only one cell, but with multiple cells it makes a problem because
different cells can have e.g. different bandwidth.

So as preparatory step for MultiRU move the code, that sets tunable
parameters for lte cells, from cell_default to cell object itself.

Rendered config changes but we can verify that the configuration, that
enb actually uses, stays the same: When enb starts it builds internal
configuration for all configured cell objects by merging cell_default +
cell's explicit configuration and using the result as the actual
configuration for the cell. There is no more cell_default in that actual
configuration. And also enb emits that actual runtime configuration in
the beginning of enb.log which we can extract and investigate(*)

So this way we can run enb with old and new generated enb.cfg and then
compare the actual configuration. It turns out to be empty:

    ~/enb# ./lteenb -n x/old/enb.cfg
    ...
    RF0: sample_rate=11.520 MHz dl_freq=1910.000 MHz ul_freq=1910.000 MHz (band 33) dl_ant=2 ul_ant=2
    (enb) quit
    ~/enb# mv log/enb.log log/enb.log.old

    ~/enb# ./lteenb -n x/out/enb.cfg
    ...
    RF0: sample_rate=11.520 MHz dl_freq=1910.000 MHz ul_freq=1910.000 MHz (band 33) dl_ant=2 ul_ant=2
    (enb) quit
    ~/enb# mv log/enb.log log/enb.log.new

    ~/enb# ../ots/json_util log log/enb.log.old >a
    ~/enb# ../ots/json_util log log/enb.log.new >b
    ~/enb# diff -u a b  # output is empty

which means that effective enb configuration before and after this patch is exactly the same.

For the reference, how internal configuration looks like is provided in the appendix.

(*) see https://tech-academy.amarisoft.com/ExtractConfigFromLog.html for details.

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1515

P.S. for now I do not plan to expose full measurement configuration in
     the generic MultiRU and so it stays in cell_default with corresponding TODO.

--------

Appendix. Dump of internal enb configuration for enb.cfg before and after the patch

```
$ ../ots/json_util log log/enb.log.old
```

```json
{
    "cell_list": [
        {
            "cell_barred": false,
            "cell_id": 0,
            "cipher_algo_pref": [],
            "cqi_period": 40,
            "cyclic_prefix": "normal",
            "dl_256qam": true,
            "dl_earfcn": 36100,
            "dpc": true,
            "dpc_pucch_snr_target": 25,
            "dpc_pusch_snr_target": 25,
            "drb_config": [ ... ],		# removed for brevity
            "ho_from_meas": true,
            "inactivity_timer": 10000,
            "initial_cqi": 3,
            "integ_algo_pref": [2, 1],
            "intra_freq_reselection": true,
            "m_ri": 8,
            "mac_config": {
                "dl_max_harq_tx": 5,
                "ul_max_harq_tx": 5
            },
            "manual_ref_signal_power": true,
            "meas_config_desc": {
                "a1_hysteresis": 0,
                "a1_report_type": "rsrp",
                "a1_rsrp": -70,
                "a1_time_to_trigger": 640,
                "a2_hysteresis": 0,
                "a2_report_type": "rsrp",
                "a2_rsrp": -80,
                "a2_time_to_trigger": 640,
                "a3_hysteresis": 0,
                "a3_offset": 6,
                "a3_report_type": "rsrp",
                "a3_time_to_trigger": 480
            },
            "meas_gap_config": "gp0",
            "n_antenna_dl": 2,
            "n_antenna_ul": 2,
            "n_id_cell": 0,
            "n_rb_dl": 50,
            "n_symb_cch": 0,
            "pdcch_format": 2,
            "pdsch_dedicated": {
                "p_a": -3,
                "p_b": -1
            },
            "phich_duration": "normal",
            "phich_resource": "1",
            "plmn_list": ["00101"],
            "prach_config_index": 4,
            "prach_freq_offset": -1,
            "pucch_dedicated": {
                "cqi_pucch_n_rb": 1,
                "n1_pucch_sr_count": 11,
                "tdd_ack_nack_feedback_mode": "multiplexing"
            },
            "pusch_dedicated": {
                "beta_offset_ack_index": 9,
                "beta_offset_cqi_index": 6,
                "beta_offset_ri_index": 6
            },
            "pusch_hopping_offset": -1,
            "pusch_max_its": 6,
            "pusch_msg3_mcs": 0,
            "q_rx_lev_min": -70,
            "rf_port": 0,
            "root_sequence_index": 204,
            "scell_list": [],
            "si_coderate": 0.2,
            "si_pdcch_format": 2,
            "si_value_tag": 0,
            "si_window_length": 40,
            "sib_sched_list": [ ... ],		# remove for brevity
            "sp_config": 7,
            "sr_period": 20,
            "srb_config": [
                {
                    "id": 1,
                    "maxRetxThreshold": 32,
                    "t_PollRetransmit": 60,
                    "t_Reordering": 45
                },
                {
                    "id": 2,
                    "maxRetxThreshold": 32,
                    "t_PollRetransmit": 60,
                    "t_Reordering": 45
                }
            ],
            "srs_dedicated": {
                "srs_bandwidth": 2,
                "srs_bandwidth_config": 2,
                "srs_hopping_bandwidth": 0,
                "srs_period": 40,
                "srs_subframe_config": 3
            },
            "tac": 1,
            "transmission_mode": 3,
            "ul_64qam": true,
            "uldl_config": 6
        }
    ],
    "com_addr": "127.0.1.2:9001",
    "enb_id": 107216,
    "gtp_addr": "127.0.1.1",
    "log_filename": "log/enb.log",
    "log_options": "all.level=error,all.max_size=0,nas.level=debug,nas.max_size=1,s1ap.level=debug,s1ap.max_size=1,x2ap.level=debug,x2ap.max_size=1,rrc.level=debug,rrc.max_size=1,phy.level=info,file.rotate=1G,file.path=/dev/null",
    "mme_list": [
        {
            "mme_addr": "127.0.1.100"
        }
    ],
    "rf_driver": {
        "args": "dev0=/dev/sdr0",
        "name": "sdr",
        "realtime": 1,
        "rx_antenna": "tx_rx",
        "tdd_tx_mod": 1
    },
    "rx_gain": 43,
    "tx_gain": 62
}
```

So that it is possible to try running enb on generated configs becuase
with previous numbers enb refuses to start:

    ~/enb# ./lteenb  x/old/enb.cfg
    ...
    x/old/enb.cfg:50: field 'dl_earfcn': range is [0:262143] (LTE Cell #0)

    ~/enb# ./lteenb  x/old/gnb.cfg
    ...
    x/old/gnb.cfg:54: invalid frequency band (NR Cell #1)

We need enb to be able to start with generated config because enb dumps in its
enb.log the real configuration that it builds from cell_list + cell_default and
nr_cell_list + cell_default - which is possible to extract via `json_util` and
which we want to check after our changes to enb.jinja2.cfg to see if we
really don't introduce semantic mistakes when e.g. moving things in between
cell_default to cell_list (see next patch).

With new frequencies enb and gnb both start ok.

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1515

For example there is no slaptap under StandaloneSlapOS and <RU>-netconf-socket
promise needs real RU hardware to be around and accepting incoming connections.

If we don't wrap such things with `if not testing` running test on testnodes will fail.

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1514

Less need to keep default synchronized and less noise in the code. We will need
to refer to `testing` more in the next patch, so it will help a bit.

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1514

Similarly to f365a440 (software/ors-amarisoft: Merge gnb.jinja2.cfg into
enb.jinja2.cfg) merge LTE and NR configuration files for UEsim into single
config as the preparation for MultiRU - there UEsim will be also handling TDD,
FDD, LTE and NR cells and UEs all at the same time.

In this patch for now we only move code without changing it and wrap parts with
`if do_lte` and `if do_nr` correspondingly.  The end result of rendered
ue-lte.cfg and ue-nr.cfg stays practically the same as shown in the appendix.

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1513

--------

Appendix. Diff for rendered ue-*.cfg before and after this patch.

```
$ ./pythonwitheggs slapos-render-config.py && git diff --no-index -w config/old/ config/out/
```

```diff
diff --git a/config/old/ue-lte.cfg b/config/out/ue-lte.cfg
index 7f1530daf..836a95e79 100644
--- a/config/old/ue-lte.cfg
+++ b/config/out/ue-lte.cfg
@@ -1,5 +1,6 @@

 #define N_RB_DL             50
+
 {
   log_options: "all.level=error,all.max_size=0,nas.level=debug,nas.max_size=1,rrc.level=debug,rrc.max_size=1,phy.level=info,file.rotate=1G,file.path=/dev/null",
   log_filename: "log/ue.log",
@@ -14,7 +15,8 @@ rf_driver: {
 },
   tx_gain: 60,
   rx_gain: 40,
-  cell_groups: [{
+  cell_groups: [
+  {
     multi_ue: true,
     cells: [
       {
@@ -27,7 +29,8 @@ rf_driver: {
     ],
     pdcch_decode_opt: false,
     pdcch_decode_opt_threshold: 0.1,
-  }],
+  },
+  ],
   ue_list: [
     {
     sim_algo: "milenage",
@@ -38,8 +41,8 @@ rf_driver: {
     impi: "",
     imsi: "001010123456789",
     K: "00112233445566778899aabbccddeeff",
-    rue_addr: "host1",
     ue_category: 12,
+    rue_addr: "host1",
     tun_setup_script: "ue-ifup",
     apn: "internet",
     }
diff --git a/config/old/ue-nr.cfg b/config/out/ue-nr.cfg
index 41758ab6f..631e1ab1e 100644
--- a/config/old/ue-nr.cfg
+++ b/config/out/ue-nr.cfg
@@ -1,3 +1,5 @@
+
+
 {
   log_options: "all.level=error,all.max_size=0,nas.level=debug,nas.max_size=1,rrc.level=debug,rrc.max_size=1,phy.level=info,file.rotate=1G,file.path=/dev/null",
   log_filename: "log/ue.log",
@@ -12,7 +14,8 @@ rf_driver: {
 },
   tx_gain: 60,
   rx_gain: 40,
-    cell_groups: [{
+  cell_groups: [
+  {
     group_type: "nr",
     multi_ue: false,
     cells: [{
@@ -26,7 +29,8 @@ rf_driver: {
       n_antenna_ul: 2,
       }
     ],
-    }],
+  },
+  ],
   ue_list: [
     {
     sim_algo: "milenage",
@@ -37,11 +41,11 @@ rf_driver: {
     impi: "",
     imsi: "001010123456789",
     K: "00112233445566778899aabbccddeeff",
-      rue_addr: "host2",
     as_release: 15,
     ue_category: "nr",
+    rue_addr: "host2",
     tun_setup_script: "ue-ifup",
     apn: "internet",
     }
-    ]
+  ],
 }
```

This is useful during development while working on ue-*.jinja2.cfg

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1513

We are going to generate more and more type of files via slapos-render-config:
currently we generate only enb.cfg and gnb.cfg, but soon it will be also
ue-*.cfg, drb_*.cfg, asn etc.

So it will be handy to easily separate generated files from their source, to
make snapshots of generated stuff for comparison, and to also e.g. to be able
to copy generated configs to a callbox/simbox for testing.

-> Move the place where generated files are emitted from config/ to config/out/

/cc @lu.xu, @tomo, @xavier_thompson, @Daetalus
/reviewed-by @jhuge
/reviewed-on nexedi/slapos!1513

Now that we have a more modern firefox, it requires secure context for
many things. This fixes `TypeError: crypto.subtle is undefined` test
failures for JIO.

See merge request nexedi/slapos!1503

Split command which is used to split pgdump files before push to gitlab backup repository crash during
backup process. This problem occurs with version starting from 9.2, older versions doesn't have this problem.
The version 9.1 is used for gitlab-backup so.

For more details, see: !1503 (diffs, comment 197515)

* master:
  stack/slapos: version up slapos.cookbook 1.0.351
  software/dufs: version up dufs 0.39.0
  *: switch to new slapos.cookbook random API for passwords
  software/dufs: include a trailing slash in published url
  software/dufs: serve index.html files instead of directory index
  software/restic-rest-server: version up v0.12.1
  component/rust: version up rustc 1.75.0
  component/apache: version up httpd 2.4.58

See merge request nexedi/slapos!1506

- switch published *arfcn types to numbers from strings
- tx_gain value was used when publishing rx_gain

I originally thought it is ORS-only, but more careful review showed that e.g.
dl_earfcn is published for both ORS and not-ORS cases:

https://lab.nexedi.com/nexedi/slapos/blob/81a83ffe/software/ors-amarisoft/instance-enb.jinja2.cfg#L224-239

* master:
  fixup! software/ors-amarisoft: Merge gnb.jinja2.cfg into enb.jinja2.cfg
  Release slapos.cookbook (1.0.351)
  random: expose hashed passwords in recipe options
  librecipe/inotify: fallback to polling when adding inotify watch fail
  software/theia/test: test that monitoring URL all have same password
  software/metabase: version up 0.48.2
  software/metabase: rename configuration parameters to kebab case
  component/postgresql: include citext extension

While not faulty in itself, f365a440 caused ors-amarisoft testsuite to start failing with

    ERROR: test_gnb_conf (testTDD.TestGNBParameters1)
    ----------------------------------------------------------------------
    Traceback (most recent call last):
      File ".../slapos-repository/software/ors-amarisoft/test/testTDD.py", line 298, in test_gnb_conf
        test_gnb_conf1(self)
      File ".../slapos-repository/software/ors-amarisoft/test/testTDD.py", line 184, in test_gnb_conf1
        self.assertEqual(conf['tx_gain'], gnb_param_dict1['tx_gain'])
    KeyError: 'tx_gain'

( https://erp5js.nexedi.net/#/test_result_module/20240119-1C2BDD94/45 )

This happens because previously gnb.jinja2.cfg emitted tx_gain/rx_gain and everything else for SDR case when `ru != "m2ru"`:

    https://lab.nexedi.com/nexedi/slapos/blob/d078060c/software/ors-amarisoft/config/gnb.jinja2.cfg#L22-55

while enb.jinja2.cfg had this only under `if bbu == "ors"`:

    https://lab.nexedi.com/nexedi/slapos/blob/d078060c/software/ors-amarisoft/config/enb.jinja2.cfg#L44-114

so since bbu != "ors" when tests are run via `testTDD.py` and `testFDD.py`
(and `testTDD-ORS.py` is not  ran at all as I explained in nexedi/slapos.toolbox!129)
this resulted in gnb.cfg not to have tx_gain/rx_gain for bbu != ors case and
corresponding test failure.

There is no real problem for ORS since for `bbu==ors` tx_gain/rx_gain are emitted.

-> Fix testsuite to pass again similarly to how it was done in b32b4a8e
   (software/ors-amarisoft: general improvement for RU (logs/promises/input
   parameters) for now).

In the future ORS case will be verified by dedicated test_ors.py which will
have checks for tx_gain and rx_gain unconditionally applied, while support for
all other cases will be handled by generic test.py which will have different
logic for how rf_driver section is generated and verified.

https://lab.nexedi.com/kirr/slapos/blob/b21d2636/software/ors-amarisoft/test/test_ors.py
https://lab.nexedi.com/kirr/slapos/blob/b21d2636/software/ors-amarisoft/test/test.py

This makes urljoin works as expected and generally makes sense because
this is a collection.