fix: allow writing to 0-dimensional arrays with sharding

[NIK-TIGER-BILL]

Closes #3751

This PR fixes a bug where writing to a 0-dimensional array with sharding enabled would crash with an IndexError in the sharding codec's vectorized chunk slice lookup.

The root cause was that get_chunk_slices_vectorized in zarr/codecs/sharding.py assumed offsets_and_lengths was at least 2-dimensional, but for 0-dimensional arrays it has shape (2,) instead of (1, 2). This caused offsets_and_lengths[:, 0] to fail with "too many indices for array".

Fix: Add an early return in get_chunk_slices_vectorized to handle the 0-dimensional case by reshaping offsets_and_lengths to (1, 2) before slicing.

Test: Added test_sharding_zero_dimensional in tests/test_codecs/test_sharding.py as a regression test.

[codecov]

Codecov Report

❌ Patch coverage is 14.28571% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.24%. Comparing base (c79425e) to head (521f425).

Files with missing lines	Patch %	Lines
src/zarr/codecs/sharding.py	14.28%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3966      +/-   ##
==========================================
- Coverage   93.27%   93.24%   -0.04%     
==========================================
  Files          87       87              
  Lines       11739    11746       +7     
==========================================
+ Hits        10950    10952       +2     
- Misses        789      794       +5     

Files with missing lines	Coverage Δ
src/zarr/codecs/sharding.py	87.50% <14.28%> (-1.60%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[chuckwondo]

Thanks @NIK-TIGER-BILL!

Would you mind also adding a test that covers your patch that handles the 0D case? The test you added shows that the original case no longer errors, but that case no longer enters the function that you patched, so your patch itself remains untested, hence the failed codecov run.