Preserve StringDType in Zarr v3 round trips

[stanbot8]

Summary

Zarr format 3 round trips change NumPy StringDType arrays to fixed-width Unicode arrays. Preserve NumPy StringDType for variables and coordinates in Zarr format 3 round trips. Closes #11466.

Design

The shared string encoder converts native StringDType arrays to fixed-width Unicode before the Zarr format 3 writer receives them. The Zarr format 3 path preserves the native stored dtype, and the shared decoder retains that dtype when the storage encoding matches it.

The implementation changes doc/whats-new.rst, xarray/backends/zarr.py, and xarray/coding/variables.py.

Tests

The baseline failed, and the candidate passed the same reproducer. The focused A/B reproducer passed on the candidate. 417 tests passed, 55 tests skipped, and 4 tests had expected failures. The repository quality checks passed in CI.

The tests are in xarray/tests/test_backends.py.

Commands

python -m pytest xarray/tests/test_coding.py xarray/tests/test_backends.py::test_roundtrip_stringdtype_zarr_v3 xarray/tests/test_backends.py::TestZarrDictStore -q
pre-commit run --files xarray/backends/zarr.py xarray/coding/variables.py xarray/tests/test_backends.py doc/whats-new.rst

Checklist

• Closes StringDType does not roundtrip through zarr #11466.
• Tests added.
• User-visible change documented in whats-new.rst.

[welcome]

Thank you for opening this pull request! It may take us a few days to respond here, so thank you for being patient.
If you have questions, some answers may be found in our contributing guidelines.