Resolve many conflicts after https://github.com/zopefoundation/ZODB/pull/357.

* master:
  Let the year float.
  Configuring for pure-python
  Specify a PyPy2 version.
  Lint the code.
  Configuring for pure-python

* Config with pure python
* Lint the code.
* Add support for Python 3.9 and 3.10.

to resolve trivial conflict on CHANGES.rst

* origin/master: (22 commits)
  Fix TypeError for fsoids (#351)
  Fix deprecation warnings occurring on Python 3.10.
  fix more PY3 incompatibilities in `fsstats`
  fix Python 3 incompatibility for `fsstats`
  add `fsdump/fsstats` test
  fsdump/fsstats improvements
  - add coverage combine step
  - first cut moving tests from Travis CI to GitHub Actions
  - ignore virtualenv artifacts [ci skip]
  tests: Run race-related tests with high frequency of switches between threads
  tests: Add test for load vs external invalidation race
  tests: Add test for open vs invalidation race
  fixup! doc/requirements: Require pygments < 2.6 on py2
  doc/requirements: Require pygments < 2.6 on py2
  fixup! buildout: Fix Sphinx install on Python2
  buildout: Fix Sphinx install on Python2
  Update README.rst
  Security fix documentation dependencies (#342)
  changes: Correct link to UnboundLocalError fsoids.py fix
  fsrefs: Optimize IO  (take 2) (#340)
  ...

Add support for Python 3.9 and 3.10.

Fix `TypeError: can't concat str to bytes` when running fsoids.py script with Python 3.

Closes #350

fsdump/fsstats improvements

Dieter Maurer notes that loadBefore cannot be deprecated yet because ZEO
essentially depends on the `end_tid` information returned by loadBefore
to update its cache:

https://github.com/zopefoundation/ZODB/pull/323#issuecomment-842021970

And to remove this dependency it would require to rework ZODB caching layer:

https://github.com/zopefoundation/ZODB/pull/323#issuecomment-845917355

So we cannot deprecate loadBefore until this rework is implemented first.

-> Remove general loadBefore deprecation, and emit loadBefore vs
loadBeforeEx warning only when actually hitting a "deletion" record,
because only that case is known to lead to data corruption.

@d-maurer says (https://github.com/zopefoundation/ZODB/pull/323#discussion_r627929959):

The changes around `loadBeforeEx` have much more impact than the `DemoStorage` fix.

--------

kirr: adjusted the text a bit: "Introduces" -> "Introduce"; add
"interface". @d-maurer, I hope it is ok.

The changelog entry uses loadBeforeEx, but we are likely to change this
name during https://github.com/zopefoundation/ZODB/pull/323 review.

loadBeforeEx uses the same docstring as loadBefore as @d-maurer suggests:

https://github.com/zopefoundation/ZODB/pull/323#discussion_r626255153

@d-maurer suggests to keep loadBefore without deprecation
(https://github.com/zopefoundation/ZODB/pull/323#pullrequestreview-650963363).

-> Don't emit warnings about deprecating loadBefore.

-> Keep the deprecation text in loadBefore interface, since loadBeforeEx
   should practically provide wider functionality without putting
   unnecessary constraint on storage implementations. In other words
   loadBefore deprecation is still there, but less aggressively
   advertised with the idea to make transition for outside-of-ZODB code
   to loadBeforeEx more smooth and with a bit more steps (we might want
   to reinstate the deprecation warnings at a later time).

Suggested by @d-maurer: https://github.com/zopefoundation/ZODB/pull/323#discussion_r625573381

@d-maurer suggests[1]:

    The ZODB logic relating to historical data (including MVCC) was largely
    centered around before. You have changed this to at - requiring wide spread
    modifications. I would much prefer to keep the before centered approach...

    (https://github.com/zopefoundation/ZODB/pull/323#pullrequestreview-650963363)

So let's change "at"-based logic to "before"-based logic and rename the new
method from loadAt to loadBeforeEx.

Define the official expansion of "ZODB"

Switch from Travis CI to GitHub Actions for testing

As suggested by @d-maurer:

https://github.com/zopefoundation/ZODB/pull/345#issuecomment-822188305
https://github.com/zopefoundation/ZEO/issues/168#issuecomment-821829116

/reviewed-on https://github.com/zopefoundation/ZODB/pull/345

For ZEO this data corruption bug was reported at
https://github.com/zopefoundation/ZEO/issues/155 and fixed at
https://github.com/zopefoundation/ZEO/pull/169.

Without that fix the failure shows e.g. as follows when running ZEO test
suite:

    Failure in test check_race_load_vs_external_invalidate (ZEO.tests.testZEO.BlobAdaptedFileStorageTests)
    Traceback (most recent call last):
      File "/usr/lib/python2.7/unittest/case.py", line 329, in run
        testMethod()
      File "/home/kirr/src/wendelin/z/ZODB/src/ZODB/tests/BasicStorage.py", line 621, in check_race_load_vs_external_invalidate
        self.fail([_ for _ in failure if _])
      File "/usr/lib/python2.7/unittest/case.py", line 410, in fail
        raise self.failureException(msg)
    AssertionError: ['T1: obj1.value (7)  !=  obj2.value (8)']

Even if added test is somewhat similar to
check_race_loadopen_vs_local_invalidate, it is added anew without trying
to unify code. The reason here is that the probability to catch load vs
external invalidation race is significantly reduced when there are only
1 modify and 1 verify workers. The unification with preserving both
tests semantic would make test for "load vs local invalidate" harder to
follow. Sometimes a little copying is better than trying to unify too
much.

For the test to work, test infrastructure is amended with
._new_storage_client() method that complements ._storage attribute:
client-server storages like ZEO, NEO and RelStorage allow several
storage clients to be connected to single storage server. For
client-server storages test subclasses should implement
_new_storage_client to return new storage client that is connected to
the same storage server self._storage is connected to.

For ZEO ._new_storage_client() is added by https://github.com/zopefoundation/ZEO/pull/170

Other client-server storages can follow to implement ._new_storage_client()
and this way automatically activate this "load vs external invalidation"
test when their testsuite is run.

Contrary to test for "load vs local invalidate" N is set to lower value (100),
because with 8 workers the bug is usually reproduced at not-so-high iteration
number (5-10-20).

/cc @d-maurer, @jamadden, @jmuchemb
/reviewed-on https://github.com/zopefoundation/ZODB/pull/345

Add test that exercises open vs invalidation race condition that, if
happen, leads to data corruption. We are seeing such race happening on
storage level in ZEO (https://github.com/zopefoundation/ZEO/issues/166),
and previously we've seen it also to happen on Connection level
(https://github.com/zopefoundation/ZODB/issues/290). By adding this test
to be exercised wrt all storages we make sure that all storages stay
free from this race.

And it payed out. Besides catching original problems from
https://github.com/zopefoundation/ZODB/issues/290 and
https://github.com/zopefoundation/ZEO/issues/166 , this test also
discovered a concurrency bug in MVCCMappingStorage:

    Failure in test check_race_open_vs_invalidate (ZODB.tests.testMVCCMappingStorage.MVCCMappingStorageTests)
    Traceback (most recent call last):
      File "/usr/lib/python2.7/unittest/case.py", line 329, in run
        testMethod()
      File "/home/kirr/src/wendelin/z/ZODB/src/ZODB/tests/BasicStorage.py", line 492, in check_race_open_vs_invalidate
        self.fail(failure[0])
      File "/usr/lib/python2.7/unittest/case.py", line 410, in fail
        raise self.failureException(msg)
    AssertionError: T1: obj1.value (24)  !=  obj2.value (23)

The problem with MVCCMappingStorage was that instance.poll_invalidations
was correctly taking main_lock with intention to make sure main data is
not mutated during analysis, but instance.tpc_finish and
instance.tpc_abort did _not_ taken main lock, which was leading to
committed data to be propagating into main storage in non-atomic way.

This bug was also observable if both obj1 and obj2 in the added test
were always loaded from the storage (added obj2._p_invalidate after
obj1._p_invalidate).

-> Fix MVCCMappingStorage by correctly locking main MVCCMappingStorage
instance when processing transaction completion.

/cc @d-maurer, @jamadden, @jmuchemb
/reviewed-on https://github.com/zopefoundation/ZODB/pull/345

buildout: Fix CI

/reviewed-on https://github.com/zopefoundation/ZODB/pull/346
/reviewed-by @d-maurer

Commit 75c703fc (Security fix documentation dependencies (#342)) changed
requirements for pygments from

    # pygments 2.6 stops the support for python2
    pygments<2.6

to

    # Silence dependabot claiming a security issue in older versions:
    pygments >= 2.7.4

However this broke CI for python2, as sphinx now fails to install:

    Collecting Sphinx
      Using cached Sphinx-1.8.5-py2.py3-none-any.whl (3.1 MB)
    ERROR: Could not find a version that satisfies the requirement pygments>=2.7.4 (from -r doc/requirements.txt (line 3)) (from versions: 0.5, 0.5.1, 0.6, 0.7, 0.7.1, 0.8, 0.8.1, 0.9, 0.10, 0.11, 0.11.1, 1.0, 1.1, 1.1.1, 1.2, 1.2.1, 1.2.2, 1.3, 1.3.1, 1.4, 1.5, 1.6rc1, 1.6, 2.0rc1, 2.0, 2.0.1, 2.0.2, 2.1, 2.1.1, 2.1.2, 2.1.3, 2.2.0, 2.3.0, 2.3.1, 2.4.0, 2.4.1, 2.4.2, 2.5.1, 2.5.2)
    ERROR: No matching distribution found for pygments>=2.7.4 (from -r doc/requirements.txt (line 3))
    The command "if [[ $TRAVIS_PYTHON_VERSION != pypy3* ]]; then pip install --upgrade --requirement doc/requirements.txt; fi" exited with 1.

(see e.g. https://travis-ci.com/github/zopefoundation/ZODB/jobs/499836548)

Try to fix that by bringing back requirement for `pygments < 2.6` on
python2.

It is not only py2 as build on py3 is also affected, e.g.

https://travis-ci.com/github/zopefoundation/ZODB/jobs/499826281

And even though Sphinx is going to be ready for newer docutils,
sphinx_rtd_theme requires docutils<0.17 for now:

https://github.com/readthedocs/sphinx_rtd_theme/blob/2026bb61ddf85a672141127b162d6f9181603725/setup.py#L119-L122

Else it is failing as:

    Got j1m.sphinxautozconfig 0.1.0.
    Getting distribution for 'sphinx_rtd_theme'.
    Got sphinx-rtd-theme 0.5.2.
    Version and requirements information containing docutils:
      Base installation request: 'Sphinx', 'docutils', 'ZODB', 'sphinxcontrib_zopeext', 'j1m.sphinxautozconfig', 'sphinx_rtd_theme'
      Requirement of sphinx_rtd_theme: docutils<0.17
      Requirement of j1m.sphinxautozconfig: docutils
      Requirement of Sphinx: docutils<0.17,>=0.12
    While:
      Installing sphinx.
    Error: There is a version conflict.
    We already have: docutils 0.17.1
    but sphinx-rtd-theme 0.5.2 requires 'docutils<0.17'.

(see e.g. https://travis-ci.com/github/zopefoundation/ZODB/jobs/499760337)

Suggested by @d-maurer.

Fix tiny issue with a remaining colon that was probably not deleted during updates of the README.

According to dependabot Pygment versions < 2.7.4 are vulnerable.
Updating to this version even though it should not be a problem with our static documentation.

Commit fc4c86e6 (Fix unbound local error when using the fsoids.py script
(#295)) wanted to refer to "issue 285", but put it as "issue 268" into
visible text.

* fsrefs: Optimize IO  (take 2)

Access objects in the order of their position in file instead of in the order
of their OID. This should give dramatical speedup when data are on HDD.

For example @perrinjerome reports that on a 73Go database it takes
almost 8h to run fsrefs (where on the same database, fstest takes 15
minutes) [1,2]. After the patch fsrefs took ~80 minutes to run on the same
database. In other words this is ~ 6x improvement.

Fsrefs has no tests. I tested it only lightly via generating a bit
corrupt database with deleted referred object(*), and it gives the same
output as unmodified fsrefs.

    oid 0x0 __main__.Object
    last updated: 1979-01-03 21:00:42.900001, tid=0x285cbacb70a3db3
    refers to invalid objects:
            oid 0x07 missing: '<unknown>'
            oid 0x07 object creation was undone: '<unknown>'

This "take 2" version is derived from https://github.com/zopefoundation/ZODB/pull/338
and only iterates objects in the order of their in-file position without
building complete references graph in-RAM, because that in-RAM graph would
consume ~12GB of memory.

Added pos2oid in-RAM index also consumes memory: for the 73GB database in
question fs._index takes ~700MB, while pos2oid takes ~2GB. In theory it could be less,
because we need only array of oid sorted by key(oid)=fs._index[oid]. However
array.array does not support sorting, and if we use plain list to keep just
[]oid, the memory consumption just for that list is ~5GB. Also because
list.sort(key=...) internally allocates memory for key array (and
list.sort(cmp=...) was removed from Python3), total memory consumption just to
produce list of []oid ordered by pos is ~10GB.
So without delving into C/Cython and/or manually sorting the array in Python (=
slow), using QQBTree seems to be the best out-of-the-box option for oid-by-pos index.

[1] nexedi/zodbtools!19 (comment 129480)
[2] nexedi/zodbtools!19 (comment 129551)

(*) test database generated via a bit modified gen_testdata.py from
zodbtools:

https://lab.nexedi.com/nexedi/zodbtools/blob/v0.0.0.dev8-28-g129afa6/zodbtools/test/gen_testdata.py

+

```diff
--- a/zodbtools/test/gen_testdata.py
+++ b/zodbtools/test/gen_testdata.py
@@ -229,7 +229,7 @@ def ext(subj): return {}
         # delete an object
         name = random.choice(list(root.keys()))
         obj = root[name]
-        root[name] = Object("%s%i*" % (name, i))
+#       root[name] = Object("%s%i*" % (name, i))
         # NOTE user/ext are kept empty on purpose - to also test this case
         commit(u"", u"predelete %s" % unpack64(obj._p_oid), {})
```

/cc @tim-one, @jeremyhylton, @jamadden
/reviewed-by @jamadden, @perrinjerome 
/reviewed-on https://github.com/zopefoundation/ZODB/pull/340

loadAt is new optional storage interface that is intended to replace loadBefore
with more clean and uniform semantic. Compared to loadBefore, loadAt:

1) returns data=None and serial of the removal, when loaded object was found to
   be deleted. loadBefore is returning only data=None in such case. This loadAt
   property allows to fix DemoStorage data corruption when whiteouts in overlay
   part were not previously correctly taken into account.

   https://github.com/zopefoundation/ZODB/issues/318

2) for regular data records, does not require storages to return next_serial,
   in addition to (data, serial). loadBefore requirement to return both
   serial and next_serial is constraining storages unnecessarily, and,
   while for FileStorage it is free to implement, for other storages it is
   not - for example for NEO and RelStorage, finding out next_serial, after
   looking up oid@at data record, costs one more SQL query:

   https://lab.nexedi.com/nexedi/neoppod/blob/fb746e6b/neo/storage/database/mysqldb.py#L484-508
   https://lab.nexedi.com/nexedi/neoppod/blob/fb746e6b/neo/storage/database/mysqldb.py#L477-482

   https://github.com/zodb/relstorage/blob/3.1.1-1-ge7628f9/src/relstorage/storage/load.py#L259-L264
   https://github.com/zodb/relstorage/blob/3.1.1-1-ge7628f9/src/relstorage/adapters/mover.py#L177-L199

   next_serial is not only about execution overhead - it is semantically
   redundant to be there and can be removed from load return. The reason
   I say that next_serial can be removed is that in ZODB/py the only place,
   that I could find, where next_serial is used on client side is in client
   cache (e.g. in NEO client cache), and that cache can be remade to
   work without using that next_serial at all. In simple words whenever
   after

     loadAt(oid, at)  ->  (data, serial)

   query, the cache can remember data for oid in [serial, at] range.

   Next, when invalidation message from server is received, cache entries,
   that had at == client_head, are extended (at -> new_head) for oids that
   are not present in invalidation message, while for oids that are present
   in invalidation message no such extension is done. This allows to
   maintain cache in correct state, invalidate it when there is a need to
   invalidate, and not to throw away cache entries that should remain live.
   This of course requires ZODB server to include both modified and
   just-created objects into invalidation messages

     ( https://github.com/zopefoundation/ZEO/pull/160 ,
       https://github.com/zopefoundation/ZODB/pull/319 ).

   Switching to loadAt should thus allow storages like NEO and, maybe,
   RelStorage, to do 2x less SQL queries on every object access.

   https://github.com/zopefoundation/ZODB/issues/318#issuecomment-657685745

In other words loadAt unifies return signature to always be

   (data, serial)

instead of

   POSKeyError				object does not exist at all
   None					object was removed
   (data, serial, next_serial)		regular data record

used by loadBefore.

This patch:

- introduces new interface.
- introduces ZODB.utils.loadAt helper, that uses either storage.loadAt,
  or, if the storage does not implement loadAt interface, tries to mimic
  loadAt semantic via storage.loadBefore to possible extent + emits
  corresponding warning.
- converts MVCCAdapter to use loadAt instead of loadBefore.
- changes DemoStorage to use loadAt, and this way fixes above-mentioned
  data corruption issue; adds corresponding test; converts
  DemoStorage.loadBefore to be a wrapper around DemoStorage.loadAt.
- adds loadAt implementation to FileStorage and MappingStorage.
- adapts other tests/code correspondingly.

/cc @jimfulton, @jamadden, @vpelletier, @jmuchemb, @arnaud-fontaine, @gidzit, @klawlf82, @hannosch