GH-45003: [Python][Docs] Update docstrings for metadata methods on Field and Schema classes

[amoeba]

Rationale for this change

The docstrings here are not completely accurate. See #45003.

I looked at whether it would make more sense or be more natural for methods like these to return an empty dict instead of None and I think None feels better from an API design and usage perspective.

What changes are included in this PR?

Updated docstrings to reflect actual behavior.

Are these changes tested?

While the changes don't need testing, I did check to see if we have unit tests for the behavior that's been documented here. We do.

Are there any user-facing changes?

Accurate docs.

• GitHub Issue: [Python][Docs] Docstring mismatch in Schema class's metadata method #45003

[amoeba]

For reviewers: I considered changing these to return an empty dict instead of None but wasn't sure what was more Pythonic or made the most sense as it relates to PyArrow. Thoughts appreciated.

[AlenkaF]

If I look for example at pa.Table.replace_schema_metadata, the metadata there can be a dict or None. So I agree with returning None here also when the metadata is empty.

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit a1627b9.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 2 possible false positives for unstable benchmarks that are known to sometimes produce them.