Fix scalar selection for nested tuple MultiIndex keys

[puneetdixit200]

Description

Closes #11341.

Full-length MultiIndex tuple labels now try exact key lookup before falling back to pandas' nested get_locs selector path. This lets tuple-valued MultiIndex levels behave as scalar keys when the full key identifies a single row, while preserving the existing fallback for nested selectors.

Tests run:

• .venv\Scripts\python -m pytest xarray\tests\test_dataarray.py::TestDataArray::test_selection_multiindex xarray\tests\test_dataarray.py::TestDataArray::test_selection_multiindex_nested_tuple_level_value xarray\tests\test_dataarray.py::TestDataArray::test_selection_multiindex_remove_unused xarray\tests\test_indexes.py::TestPandasMultiIndex::test_sel -q
• .venv\Scripts\python -m pytest xarray\tests\test_indexes.py::TestPandasMultiIndex -q
• .venv\Scripts\python -m pytest xarray\tests\test_dataarray.py -q -k multiindex
• .venv\Scripts\python -m ruff format --check xarray\core\indexes.py xarray\tests\test_dataarray.py
• .venv\Scripts\python -m ruff check xarray\core\indexes.py xarray\tests\test_dataarray.py
• git diff --check

Checklist

• Closes A single nested tuple MultiIndex key is located correctly but preserves the dimension #11341
• 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-assisted content.
  • I have tested any AI-assisted content in my PR.
  • I take responsibility for any AI-assisted content in my PR.
    Tools: OpenAI language model assistance.