Fix jaxtyping>=0.3.0 support for NumPy annotations

[pablovela5620]

Fixes #643.

Summary

I used Codex to help investigate the regression, isolate the root cause, and validate the fix.

pyserde's jaxtyping support broke with jaxtyping>=0.3.0 because jaxtyping annotations no longer subclass their backing array type.

Before 0.3.0, an annotation like jaxtyping.Float[np.ndarray, "2 2"] was a subclass of np.ndarray, so pyserde detected it with issubclass(..., np.ndarray).

From 0.3.0 onward, jaxtyping annotations inherit from jaxtyping.AbstractArray instead, while still exposing array_type. That caused pyserde's jaxtyping NumPy handling to stop matching and fall through to SerdeError: Unsupported type: ndarray.

This patch keeps the fix small:

• switch jaxtyping NumPy detection to jaxtyping.AbstractArray plus array_type is np.ndarray
• remove the jaxtyping<0.3.0 pin
• add a focused regression test for the detection helper
• keep the existing jaxtyping round-trip tests as the main behavior check
• update the standalone jaxtyping example to use concrete NumPy dtypes accepted by current NumPy

Local validation

Ran locally:

make test
make examples
uv run pytest tests/test_numpy.py -k jaxtyping -v
uv run python examples/type_numpy_jaxtyping.py
uv run ruff check serde/numpy.py tests/test_numpy.py examples/type_numpy_jaxtyping.py

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.81%. Comparing base (f09d503) to head (9b0a9c4).
⚠️ Report is 65 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #730      +/-   ##
==========================================
- Coverage   90.83%   90.81%   -0.03%     
==========================================
  Files          13       13              
  Lines        2040     2307     +267     
  Branches      370      449      +79     
==========================================
+ Hits         1853     2095     +242     
- Misses        124      141      +17     
- Partials       63       71       +8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[yukinarit]

LGTM, Thanks!