This reverts commit 2c5c1147.

If this command fail, instanciation should not succeed.

If this command fail, instanciation should not succeed.

This enable vieiwing applications logs directly in grafana, as long as
loki agent can access the log file and the regular expressions to parse
the logs are defined

Other changes:
 - provision data sources automatically so that user do not have to add
   them manually
 - update TODO in README
 - introduce a SR test
 - switch to dep to install dependencies, because that's how these
  packages manage their dependencies.

Versions up:
  grafana: v6.3.0
  telegraf: v1.11.1-0
  influxdb: v1.7.8rc1
New:
  loki: v0.3.0

This runs in a go workspace, but we don't need to always have the
(gowork) marker.

We don't want theia to use /tmp/ this prevents running more than one
instance (for example we can only install vscode extensions once. golang
language server is also broken)

if yarn.lock is changed, part should be re-installed. This is only
useful during development.

- exclude .pyc in files list
- exclude .env for virtualenv
- don't watch git (not sure it's really needed)
- use 2 spaces on editors
- use Vifib's plantuml service
- use Ctrl for multi cursor (for chromeos where alt is right click)

Also stop including python, as we can use vscode python instead.
Add a README to describe how to do this.

(our latest version)

I set that to try to be independent from user config files/environment
variables, but that was a bad idea and not needed (I did that to
investigate an issue I had with jedi, but it was not depending on
environment variables - it was depending on the presence of a
buildout.cfg in current directory)

json was not formatted with `./format-json` this was causing test
failures.

The main change is a new and unique way to override parameters depending on where these parameters are used:
- for NEO, the goal is to avoid a huge `node_list` value (in particular with MyRocks settings), whereas most of the time all values in the list are the same
- for ERP5, we need a way to define different ZODB cache settings depending on the zope id (e.g. huge cache for some specialized zopes, a small cache for the validation node, etc. whereas currently a huge cache for all nodes would waste a lot of RAM)

For any `key=value` parameter that can be overridden, overriding is done as follows:
- `key` defines the default value (not required if the SR already defines one)
- `key!=[[pattern, value],...]`: the SR takes the value of the first 2-tuple whose *some reference* matches the pattern

For NEO:
- new `node_count` and `node` parameters
- `node!` can be used to override `node`: the *reference* is the partition reference (`node-<i>` for standalone NEO, and `neo-<i>` for ERP5)
- `node_list` is deprecated (if still passed, the new parameters must not be present)

For ERP5 ZODB settings, parameters can be overridden at both:
- zodb level (e.g. `pool-size`, `cache-size`... oh, I've just discovered there exists `cache-size-bytes`)
- storage level (e.g. `compress`, `cache-size`, etc. in the case of NEO)

Here, the *reference* is the zope id (e.g. `actitivies-0`)

The `pattern` is a Python regex that must match the whole reference.

Another not-so-small change is that the schema is fixed to only require NEO `cluster` parameter when instanciating a standalone NEO.

And a few fixes.

/reviewed-on nexedi/slapos!661

Some eggs to provide "intelliense" for python withing ERP5

/reviewed-on nexedi/slapos!649