Fix deprecation of setting shape on a numpy array in 2.5+

[mscheltienne]

Fixes deprecation observed in numpy 2.5+ (pre-release):

 .venv/lib/python3.13/site-packages/zarr/codecs/vlen_utf8.py:98: in _decode_single
    decoded.shape = chunk_spec.shape
    ^^^^^^^^^^^^^
E   DeprecationWarning: Setting the shape on a NumPy array has been deprecated in NumPy 2.5.
E   As an alternative, you can create a new view using np.reshape (with copy=False if needed).

[codecov]

Codecov Report

❌ Patch coverage is 50.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 60.90%. Comparing base (acdd892) to head (2bc0cbc).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/zarr/codecs/vlen_utf8.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3648   +/-   ##
=======================================
  Coverage   60.90%   60.90%           
=======================================
  Files          86       86           
  Lines       10174    10174           
=======================================
  Hits         6196     6196           
  Misses       3978     3978           

Files with missing lines	Coverage Δ
src/zarr/codecs/vlen_utf8.py	28.07% <50.00%> (ø)

🚀 New features to boost your workflow:

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

[d-v-b]

thank you!

[mscheltienne]

@d-v-b Note that this change is not fully backward compatible, if we want it backward compatible we would set copy=False instead of the default copy=None. I consider that copy=None is an enhancement as it will now:

• (1) Try to reshape without copy (view)
• (2) If it can't, reshape with a copy

If we set copy=False, it will explicitly raise if the view is not possible. Up-to-you, let me know if you think this added fallback mechanism is an enhancement or if hiding this copy is detrimental.

[d-v-b]

As decoded is an object array, when would we not be able to create a view?

[mscheltienne]

Well it should- and personally I don't see the harm in adding this fallback- but if you prefer to have an explicit raise instead to catch eventual future changes that would somehow yield a non contiguous array of objects, then setting copy=False would be better.

[d-v-b]

in that case the copy=False option seems good. Probably better to err on the side of avoiding hidden allocations wherever possible.