tests: fixes all warnings

[noirbizarre]

Description

This PR fixes all raised warnings during tests (making it easier to read actual errors).

Moving from 30 warnings and a duration of ~40s:

to no warnings for a duration of less than 10s:

The time gained is because we don't rely on sleep anymore.

Checklist

• I have read the contributing guidelines

Code Changes

• ignore our own expected deprecation warnings in pyproject.toml
• Introduce a time-controlled util: UtilFixture helper (freeze time with FreezeGun once required):
  • exposes all git operations in a time-controlled manner (each git operation advance time by 1s).
  • exposes a cli(*args) helper also time-controlled (underlying git operations have the time frozen too). Reduce the amount of boilerplate to execute the cli during tests
• use this fixture instead all imported git helpers, including wait_for_tag in test_changelog_command.py (the only one still depending on wait_for_tag)
• add missing type annotation in test_changelog_command (all but those based on legacy pytest tmpdir/py.path.local type which is badly typed)
• update regression files depending on an invalid mocked date (1970-01-01)

The fixtures has not yet replaced helpers in other tests to reduce the size of the PR (will be done in separate pull requests)

Documentation Changes

None

Expected Behavior

Faster (no sleep) and fully deterministic (full control over time) tests execution.
Cleaner output (no more pages of warnings)
Easier to maintain tests

Steps to Test This Pull Request

Execute poetry test on master and this branch to observe the differences of execution (speed and output).

Additional Context

This was a long task but this was also (IMHO) required to make it easier to maintain tests on long term.
I'll update other test files when this one is merged

Fixes #1623

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.88%. Comparing base (548621c) to head (77c20c7).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1746      +/-   ##
==========================================
- Coverage   97.92%   97.88%   -0.04%     
==========================================
  Files          60       60              
  Lines        2604     2604              
==========================================
- Hits         2550     2549       -1     
- Misses         54       55       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Lee-W]

This PR is awesome! Just browsed through it Left a few nits and need some time to take a deeper look

[woile]

Beautiful! Thanks. LGTM!

[noirbizarre]

PR updated with all changes 👍🏼

[bearomorphism]

qq: will this test fail tomorrow?

[bearomorphism]

I just wanted to know where 2025-12-29 comes from. I haven't found related logic besides datetime.now() in this PR.

[bearomorphism]

Let me quickly rerun a test, it's 2025-12-30 UTC time now.

[noirbizarre]

No it won't fail, time is frozen with freezegun (I just rebased the PR on master and CI is still passing while it's 2025-12-30).

I just set yesterday date as freeze time because I add to set one and I though freezing on the day of change give an hint on when it has been done.

The test is decorated with @pytest.mark.freeze_time('2025-12-29'), so time has been frozen to 2025-12-29T00:00:00 at the start of the test and each call to util.*() methods will advance time by 1s. It makes it easy to manage changelog dates in test while also guaranteeing commit and tags sequentiality without ever waiting with sleep().
You can also manually advance time with util.tick().

When I'll port the other tests, if some require bigger time intervals, I'll add another util.move_to(date) method but it was not the case for those tests.