Summary: Related to: T3432

Reviewers: #reviewers

The default implementation is a naive 'for' loop calling 'get'
sequentially.

This replaces this default implementation with a local asyncio loop,
which manages a task for each of the blobs to fetch.

To avoid code duplication, this also rewrites 'get' to use the same
code as 'get_batch'. Quick benchmarks did not show a noticeable
slowdown when doing this.

A future commit will start using aio.ContainerClient instances, which must
be created on the fly from these URLs.

Enable to check package documentation can be built without producing
sphinx warnings.

The sphinx environment is designed to be used in continuous integration
in order to prevent breaking documentation build when committing changes.

The sphinx-dev environment is designed to be used inside a full swh
development environment.

Related to T3258

Related to T2265

This makes get_objstorage consistent with most of the other `get_foo` factories
available in other swh modules.

We will need to do a pass to drop the few hundred `args` that are used
currently.

Related to T2610

Related to T2610

flake8 hook has been removed from https://github.com/pre-commit/pre-commit-hooks
so now use the one from https://gitlab.com/pycqa/flake8

by moving import statements in functions and stop importing
swh.objstorage.factory.* in swh.objstorage.

Warning: this later breaks a few dependencies that had not updated
the way they import get_objstorage.

Related to T2575.

pytest wastes a lot of time in .hypothesis and .git; this commit excludes them.

Related to T2105

If an objstorage is read-only, the add() method will return None. The
multiplexer should return one of the non-None values if it finds one, instead of
the (inconsistent) first value returned by any of the threads.

There's an unavoidable race condition in the azure objstorage between the
existence check and upload. When this race condition happens and we upload the
same blob twice, one of the uploads is rejected with a `ResourceExistsError`.

This change brings the azure objstorage back to idempotence.

Fixes sentry error SWH-STORAGE-G6.

This fixes the ci build [1]

[1] https://jenkins.softwareheritage.org/job/DOBJS/job/tests/663/console

The new version is what Debian unstable is carrying, so use that. Unfortunately
the situation with respect to the old API hasn't been clarified by upstream yet.

https://github.com/Azure/azure-storage-python/issues/655

This is needed to "empty" the swh.objstorage's *module* from any
non-subpackage objects, so we can implement the swh.objstorage.replayer
outside the swh.objstorage's repo, in a dedicated repo/project.

Related to T2367

- blackify all the python files,
- enable black in pre-commit,
- add a black tox environment.

Related to https://github.com/Azure/azure-storage-python/issues/437

Meaning that:
* server should raise 400 on these errors instead of 500
* client should catch them and re-raise them.

This allows to store objects alongside other data in a given S3 bucket, which is
what we do in our current Amazon S3 export.