GH-48759: [Python] Remove Python 3.0.0 condition for pandas/issues/50127

[HyukjinKwon]

Rationale for this change

Tests test_binary_column_name and test_roundtrip_with_bytes_unicode were being skipped on pandas 2.0+ due to a pandas regression (pandas-dev/pandas#50127) that caused NotImplementedError: |S3 when roundtripping DataFrames with bytes column names.

This issue was already fixed in commit e2ac52d (GH-44171) by changing an if statement to elif in _reconstruct_columns_from_metadata(). The fix prevents bytes-typed columns from falling through to the level.astype(dtype) call that triggered the pandas error.

What changes are included in this PR?

Remove the skip conditions from two test cases.

Are these changes tested?

Yes all related tests pass.

Are there any user-facing changes?

No.

• GitHub Issue: [Python] Remove Python 3.0.0 comparison for pandas/issues/50127 #48759

[github-actions]

⚠️ GitHub issue #48759 has been automatically assigned in GitHub to PR creator.

[raulcd]

@WillAyd @jorisvandenbossche any thoughts on this?

[WillAyd]

I am fine with it. Looks like a pretty old issue - I don't think its worth having to move the upper version bound here until this gets fixed

[HyukjinKwon]

Could this be merged by any change? 🙇

[raulcd]

I am confused, shouldn't we have seen this failing on the pandas-devel and pandas upstream job as those are 3.0.0 dev already? This specific test doesn't fail for those, I am wondering whether this could have been solved as part of a different issue.

[HyukjinKwon]

Actually let me check.

[HyukjinKwon]

huh? tests passed but pandas-dev/pandas#50127 is still not resolved

[HyukjinKwon]

Alright. e2ac52d accidentally fixed this issue together. It does not trigger level.astype(dtype) anymore which would fail per pandas-dev/pandas#50127.