Bump development tools: mypy, codespell, isort, ...

Move all tools configuration in pyproject.toml.

Remove no longer needed mypy overrides.

We already log a final failure, having failures that will be retried get
logged, especially at ERROR level, is confusing when glancing at logs.

Some recent changes in the tenacity package broke the tests
in the test_replay_errors file, so adapt the implementation
of the copy_object_q function to make them pass again.

This will let us do a "passive" scrub of the source objstorage before
populating the destination objstorage. This will also avoid replicating
the "wrong" object when replicating an object subject to a sha1 collision.

This will allow reusing the error logging outside of a retried function.

They were previously stored in binary format making them not
exploitable from the sentry web UI.

This reverts commit ebf20c2d.

The actual fix for this is to make swh.objstorage an actual namespace
instead of a package, aka remove the (empty) __init__.py file in
swh.objstorage.

It seems this "package used a s a namespace" used to work, but not any more.

As swh.objstorage.replayer is a sub-package in the swh.objstorage
implicit namespace, pytest could not import the swh.objstorage.replayer
modules as it could not find the swh.objstorage.replayer one.

To fix that issue and ensure than tests can be run in a local venv
but also with tox, remove the use of the `--import-mode=importlib`
pytest option and use relative imports instead of absolute ones
when required.

Related to swh/meta#5075.

Stop relying on the implicit assumption that object storages can be
accessed with bare sha1 hashes.

Keep the sha1-exclude-file mechanism as the S3 inventory for which it
was introduced still relies on sha1, at least for now.

To pave the way for future changes in swh.objstorage, also adapt tests
to remove the assumption that InMemoryObjStorage state will remain keyed
by sha1 (by using the _state_key function).

This changes the key in the redis failure reporter to reference all hashes
instead of just sha1.

Convert README from markdown to ReST to make it embeddable in
docs/index.rst

It now needs types-click which is indeed a dependency of
swh.core[testing].

- add config entry in pyproject.toml
- add a section in tox.ini (not executed by default for now)
- fix test_cli.py to please ruff (C401 rule)

and classify as Production/Stable.

this allows to make it actually benefit from multithreading, by not
sharing objstorage instances between threads.

Note: the doctest formely in process_replay_objects_content's docstring
has been moved to a proper test in test_replay.py; the required mocking
to make it work is not suitable for docstring anymore.

Related to swh/meta#4960

Related to swh/meta#4959

This fixes python 3.7 support due to poetry, a dependency of isort, that
removed support for that Python version in a recent release.

In order to remove warnings about /apidoc/*.rst files being included
multiple times in toc when building full swh documentation, prefer to
include module indices only when building standalone package documentation.

Also include them the proper sphinx way.

Related to T4496

This option allows to set a maximum size for content objects to be
replicated in the destination objstorage.

Add a few debug logging statements and improve the info-level summary
statement at the end of `process_replay_objects_content()`.

Fix tests accordingly.

to make it a bit more robust and give more details on what's going on in
case of errors.

Add tests for those error situations.

Side effect of rDJNL1d879f1dd6247b74d7e58dd0b4bd7b498ab506f0.

- pre-commit from 4.1.0 to 4.3.0,
- codespell from 2.2.1 to 2.2.2,
- black from 22.3.0 to 22.10.0 and
- flake8 from 4.0.1 to 5.0.4. Also freeze flake8 dependencies.

Also change flake8's repo config to github (the gitlab mirror
being outdated).