[HELP WANTED] [SEND PRs] Make use of pytest features when running tests

[webknjaz]

We've switched over the test runner to pytest and there are a few tests
that already use it natively too. Still, the transition isn't over yet, it's an ongoing effort and we ask for the help of the community to get this over with faster.

Help wanted

Here's how you can help:

1. When adding new test modules (or just test functions), never import unittest. Follow, pytest conventions, use fixtures. Use test functions, classes are unnecessary here.
   Use @pytest.mark.parametrize to have the same test body run against multiple inputs: each one of them will show up in the report separately. Use assert statement, don't inherit from unittest.TestCase. Use built-in tmp_path fixture (not tmpdir — it's deprecated) to work with temporary data on disk and make sure that multiple tests don't hit the resources at the same time.
2. When reporting a bug, write an integration test as a new pytest test function. It should be failing. Cover as many edge and negative cases as you can.
   Send a PR demonstrating the red CI. After that, add one separate commit on top of that in the PR branch adding @pytest.mark.xfail as per https://pganssle-talks.github.io/xfail-lightning. Use raises= arg if it's not the assertion that's failing. Follow the rules from (1).
3. Send us separate PRs migrating legacy unittest-based modules to pytest. Follow the rules from (1).

pytest-based test module example

https://github.com/ansible/ansible-lint/blob/065faa5/test/TestIncludeMissingFileRule.py

[webknjaz]

UPD: #945 is going to help with enforcing the use of pytest over unittest. If one were to send PRs updating the ignored modules, they'll need to drop the ignore-comments from the top.