this reduces logic duplication, YAY!

This way mypy is more thorough in finding source code files to check.
Also, this is consistent with how we invoke other linters, e.g., flake8.

this is needed as individual packages might need specific plugins
(e.g., for very dynamic libraries like django, sqlalchemy, etc.) and
there is no guarantee those plugins will always be compatible

plus, per-package configuration is what we are already doing for other
test/lint tools, e.g., tox, pytest, etc. Diverging for mypy does not
seem a great idea

this way "make typecheck" could work individual packages even when they are not
installed out of swh-environment---which would be the case for CI runs, for
instance

Summary:
Also port the coverage flags to be compatible with pytest-cov.

Kinda sorta close T1261.

Test Plan: looks like make test is happy

Reviewers: douardda

Maniphest Tasks: T1261

Differential Revision: https://forge.softwareheritage.org/D569

Reviewers: olasd, #reviewers, douardda

Reviewed By: #reviewers, douardda

Subscribers: douardda

Differential Revision: https://forge.softwareheritage.org/D498

Run codespell only for:

  - Python files related to the swh module

  - swh module documentation source files

  - swh module README file

leave the decision to Makefile.sphinx

Note that actual linting still does not work, because of pylint issues with PEP
420 namespaces. Still, the build toolchain scaffolding is now in place.

Summary:
Using python3 -m nose instead of the nosetests3 entrypoint allows
us to call the nosetests that depends of the python version we chose instead of
one that might not be up to date or present at all.

Test Plan: make check

Reviewers: #reviewers!, olasd, zack, ardumont

Differential Revision: https://forge.softwareheritage.org/D120

Which supports files that were not imported during tests too.

to clean up Python byte code files, if needed

this avoids failing on "yield from"

to work around current nose issues with namespace packages

e.g., swh-loader-debian