GH-31318: [Python] Add fixed-offset timezones to Hypothesis test strategy

[alex-anast]

Rationale for this change

The Hypothesis test strategy for timezones (pyarrow.tests.strategies.timezones) only generates named IANA timezones (via pytz and zoneinfo). It does not generate fixed-offset timezones like +05:30 or -03:00. As noted in #31318, this is not supported out of the box by Hypothesis, so a custom strategy is needed. Fixed-offset timezones are already used in manual tests and handled correctly by Arrow, but are never exercised via property-based testing.

What changes are included in this PR?

Adds a fixed_offset_timezones strategy to python/pyarrow/tests/strategies.py that generates datetime.timezone objects with offsets ranging from UTC-12:00 to UTC+14:00, with minute components of 0, 30, or 45 (covering all real-world UTC offsets). This strategy is included in the existing timezones strategy across all dependency configurations.

Are these changes tested?

This change improves test infrastructure itself -- it extends the timezones strategy used by existing Hypothesis-based tests. The generated fixed-offset timezones were verified to work correctly with pa.timestamp().

The test command used was:

export MAMBA_ROOT_PREFIX=$PWD/.micromamba && eval "$(/tmp/bin/micromamba shell hook -s bash)" && micromamba activate pyarrow-dev && export  LD_LIBRARY_PATH=$CONDA_PREFIX/lib:$LD_LIBRARY_PATH && python -m pytest pyarrow/tests/test_scalars.py pyarrow/tests/test_array.py -v -k test_dunders 

Are there any user-facing changes?

No. This only affects the internal test suite.

AI-generated code disclosure

This PR was developed with assistance from an AI coding tool (Claude, Anthropic). All changes have been reviewed, understood, and verified.

• GitHub Issue: [Python] Update timezones strategy to include fixed offsets #31318
  Closes [Python] Update timezones strategy to include fixed offsets #31318

[github-actions]

⚠️ GitHub issue #31318 has been automatically assigned in GitHub to PR creator.

[github-actions]

⚠️ GitHub issue #31318 has been automatically assigned in GitHub to PR creator.

[raulcd]

@github-actions crossbow submit test-conda-python-3.11-hypothesis

[github-actions]

Revision: 6c9a20e

Submitted crossbow builds: ursacomputing/crossbow @ actions-16091e66e1

Task	Status
test-conda-python-3.11-hypothesis

[AlenkaF]

The extended hypothesis failure is currently unrelated, opened an issue for it here: #49846

[AlenkaF]

Thanks for the PR!
The only comment I have for now is that the current strategy can produce UTC out of the valid UTC range which is UTC-12 and UTC+14, for example UTC-12:45. Datetime will accept it but as this is invalid IANA timezone it should be improved, it also might produce errors in other parts.

[github-actions]

⚠️ GitHub issue #31318 has been automatically assigned in GitHub to PR creator.

[github-actions]

⚠️ GitHub issue #31318 has been automatically assigned in GitHub to PR creator.

[github-actions]

⚠️ GitHub issue #31318 has been automatically assigned in GitHub to PR creator.

[alex-anast]

@AlenkaF thanks for the review!
You're right. I've pushed a follow-up that enforces both bounds, making sure that no offset fails outside the valid UTC-12 to UTC+14 range. Can you please re-check?

edit: on hindsight, a lower bound filter is unnecessary since the lower bound can never be violated -- harmless, clear and defensive, but dead code.

[AlenkaF]

Thanks for the update! There is a PR up to fix the hypothesis build: #49847. After that is merged you can rebase and we can run the build here, to double check the logic.