Fixes for test_unique narwhals test

[scott-routledge2]

Changes included in this PR

• adds extra projection in "subset" case to ensure column ordering matches Pandas
• adds error checking for duplicated in the JIT fallback path

Testing strategy

narwhals: test_unique -- all params now passing, (keep="none" falls back to pandas)

User facing changes

Checklist

• Pipelines passed before requesting review. To run CI you must include [run CI] in your commit message.
• I am familiar with the Contributing Guide
• I have installed + ran pre-commit hooks.

[codecov]

Codecov Report

❌ Patch coverage is 37.50000% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 68.79%. Comparing base (c33fbb5) to head (a90b4bb).
⚠️ Report is 115 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #920      +/-   ##
==========================================
+ Coverage   66.68%   68.79%   +2.10%     
==========================================
  Files         186      195       +9     
  Lines       66795    67557     +762     
  Branches     9507     9594      +87     
==========================================
+ Hits        44543    46474    +1931     
+ Misses      19572    18251    -1321     
- Partials     2680     2832     +152     

[scott-routledge2]

The results don't seem to be deterministic with 3 workers. The original dataframe is: {'a': [1, 3, 2], 'b': [4, 4, 6], 'z': [7.0, 8.0, 9.0]} Sometimes df.unique(subset=["b"], keep="any") is
{'a': [1, 2], 'b': [4, 6], 'z': [8.0, 9.0]} (z and a values don't match), and sometimes it's correct.

Edit: oh this could be related to how dataframes are evaluated. The correct result is either going to be (1, 7.0) or (3, 8.0) but if the columns are evaluated independently then that would explain the sometimes mismatch.

[ehsantn]

Looks good to me.

[DrTodd13]

LGTM! Thanks.