Return a copy from get_index

[chiruu12]

Fixes #2949.

get_index returned the stored pandas index. pandas allows .name to be set in place, so a caller could rename the coordinate of the array or dataset they got the index from:

array = xr.DataArray(np.zeros(2), coords={"x": ["a", "b"]}, dims=["x"])
array.get_index("x").name = "renamed"
array.get_index("x").name   # 'renamed' before this change, 'x' after

Dataset was affected the same way, since get_index lives on DataWithCoords.

The change is a copy(deep=False) on the way out. #3126 proposed the same fix in 2019 and stalled on a question about whether it would cost anything, which nobody answered. It does not. A shallow copy of a pandas index is constant time, since the object is immutable and the copy shares the same array:

      1,000 elements   1.16 us
  1,000,000 elements   1.15 us
 10,000,000 elements   1.19 us

pandas 3.0.5, mean of 10000 calls. The comment at xarray/core/variable.py:3005 already states this.

Scope note. Indexes.__getitem__ still returns the stored object, so ds.indexes["x"] is unchanged. xarray/tests/test_indexes.py:630 asserts that identity, so copying there is an API decision rather than a bug fix, and I left it alone. to_index() and to_dataframe() were already safe because they route through set_names().

Tests: one in test_dataarray.py and one in test_dataset.py. Both fail on main and pass here.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.