Fix idxminmax for interval coordinates

[lavaFreak]

Description

Fix DataArray.idxmax() and DataArray.idxmin() when the reduced dimension uses interval coordinates. _calc_idxminmax() was replacing the coordinate's original pandas index adapter with a raw extension array before the final label lookup, which causes scalar interval selections to fail with TypeError: len() of unsized object.

This keeps pandas extension-backed coordinates on their original adapter for the final lookup, adds a regression test for both idxmax and idxmin, and documents the bug fix in doc/whats-new.rst.

Checklist

• Closes .idxmax() fails if coordinates are intervals #11300
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.
    Tools: Codex

[dcherian]

i suspect the real solution is to fix to_like_array to handle extension arrays?

[lavaFreak]

Good call. I moved the fix into to_like_array() in d0f22b94 instead of keeping the one-off special case in _calc_idxminmax().

For NumPy-backed lookups, to_like_array() now converts allowed pandas extension arrays to a NumPy object array, which fixes the scalar interval selection path while keeping the idxmin/idxmax regression green.

Validation I reran locally:

• .venv/bin/python -m pytest -o addopts= xarray/tests/test_dataarray.py -k 'test_idxminmax_interval_coords or test_idxmax or test_idxmin'
• .venv/bin/python -m ruff check xarray/computation/computation.py xarray/compat/array_api_compat.py xarray/tests/test_dataarray.py