[SPARK-56765][PYTHON] Suppress pyarrow.compute mypy attr-defined errors

[201573]

What changes were proposed in this pull request?

This PR adds narrow # type: ignore[attr-defined] annotations for pyarrow.compute calls whose symbols are missing from the PyArrow 24.0.0 stubs: floor_temporal, assume_timezone, and local_timestamp.

Why are the changes needed?

PyArrow 24.0.0 still exposes these compute kernels at runtime, but its type stubs do not expose them to mypy. This currently breaks PySpark lint with attr-defined errors.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

• PYTHON_EXECUTABLE=/tmp/spark-56765-mypy-venv/bin/python ./dev/lint-python --compile
• /tmp/spark-56765-mypy-venv/bin/ruff format --check python/pyspark/sql/pandas/types.py python/pyspark/sql/conversion.py
• git diff --check
• Targeted mypy with pyarrow==24.0.0 confirmed these pyarrow.compute attr-defined errors are suppressed.

Was this patch authored or co-authored using generative AI tooling?

Generated-by: OpenAI Codex (GPT-5)

[LuciferYang]

cc @zhengruifeng as discussed offline, the daily test failures on branch-4.x/branch-4.2/branch-4.1 are likely related to this issue.

[devin-petersohn]

LGTM

[gaogaotiantian]

I think the solution in #55744 is better. We should try to avoid ignoring unused-ignore. This is an issue for pyarrow.compute and we should just ignore that specific module.

[LuciferYang]

+1, agree with @gaogaotiantian

[201573]

Thanks for the feedback. I agree that #55744 is cleaner since it keeps the suppression scoped to pyarrow.compute in mypy.ini instead of adding unused-ignore at the call sites.

I'll close this in favor of #55744.