fix: allow require_array to accept a ZDType

[JOhnsonKC201]

Summary

Closes #3377.

Group.require_array could not accept a Zarr data type, even though every other
array-creation method on the group already does. Requiring an array that exists
raised:

TypeError: Cannot interpret 'Int32(endianness='little')' as a data type

AsyncGroup.require_array was annotated npt.DTypeLike | None and normalised
its argument with np.dtype(dtype), which cannot consume a ZDType. Its
siblings (create_array and friends) take ZDTypeLike. So the "array already
exists" branch always blew up for a ZDType, while the string and NumPy forms
worked.

The fix widens the annotation to ZDTypeLike | None and normalises with
parse_data_type(...).to_native_dtype(). AsyncArray.dtype already returns the
native dtype, so the exact / np.can_cast comparisons below are unchanged.

Two follow-on details: parse_data_type(None) resolves to float64, which is
exactly what np.dtype(None) gave, so the default path is unchanged; and
removing the last real use of numpy.typing left the import unused, so it is
dropped (the three remaining npt. mentions in this file are in docstrings).

For reviewers

The bit worth a second look is the None case, since np.dtype(None) silently
means float64 and I wanted to preserve that rather than start rejecting it. I
checked all four input forms round-trip to the same array dtype: ZDType,
np.dtype, "int32", and None.

I am confident about the rest: it is a normalisation swap, and behavior for
string and NumPy dtypes is identical.

Author attestation

• I am a human, these are my changes, and I have reviewed and understood every change and can explain why each is correct.

TODO

• Add unit tests and/or doctests in docstrings
• Add docstrings and API docs for any new/modified user-facing classes and functions
• New/modified features documented in docs/user-guide/*.md
• Changes documented as a new file in changes/
• GitHub Actions have all passed
• Test coverage is 100% (Codecov passes)

Test plan

• tests/test_group.py::test_require_array_zdtype, covering ZDType,
  np.dtype and str. It fails on main with the TypeError above.
• tests/test_group.py: 529 passed, 15 skipped.
• tests/test_api.py tests/test_array.py: 1654 passed, 44 skipped.
• ruff check and ruff format --check clean.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.00%. Comparing base (bd3e398) to head (a5f946d).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4189   +/-   ##
=======================================
  Coverage   94.00%   94.00%           
=======================================
  Files          91       91           
  Lines       12791    12791           
=======================================
  Hits        12024    12024           
  Misses        767      767           

Files with missing lines	Coverage Δ
src/zarr/core/group.py	95.33% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[d-v-b]

can you parametrize this test over (input, expected output) pairs, and ensure that the dtype=None case is included? Pairs like [("int32", Int32()), (None, Int64()]

[d-v-b]

correction: the pair for the None case would be (None, Float64())

[d-v-b]

hi @JOhnsonKC201 is it OK if I push some changes to your branch to get some of the remaining issues resolved? we just need a few changes before this can be merged

[JOhnsonKC201]

Sure, go ahead.

I just pushed 98376b1, which parametrizes test_require_array_zdtype over (input, expected) pairs with the (None, Float64()) case included, since that looked like the open review comment. It landed a few minutes after you asked, so it may not have been there when you looked.

Happy for you to push whatever else is needed to get this over the line.

[d-v-b]

thanks @JOhnsonKC201!