TST: add tests for nullable astype behavior (NaN/<NA> preservation)

[AnonToky]

Adds small, focused tests covering nullable dtype casting behavior to ensure missing values are preserved or converted as expected during astype operations.

Adds pandas/tests/dtypes/test_nullable_astype.py with:

test_series_astype_nullable_int_preserves_nans: float -> Int64 should convert NaN to .
test_series_astype_from_nullable_int_to_float_roundtrip: Int64 -> float -> Int64 roundtrip preserves missingness.
test_nullable_series_astype_various_dtypes_preserve_missing: parametrized coverage for Int32/Int64/Float32/Float64.

Rationale:

Small, deterministic tests that exercise edge cases around pd.NA and numpy NaN during dtype conversion.
No behavioral changes to core code; this enhances test coverage and guards against regressions.
How to run locally: pytest -q pandas/tests/dtypes/test_nullable_astype.py

Notes:

Uses pandas.testing assertions and pytest parametrization.
No external dependencies or network I/O.

[jbrockmendel]

This looks like AI.

[mroeschke]

Thanks for the PR, but it doesn't appear to be related to an open issue and doesn't follow the testing conventions of the repository so closing