Keep integers on the python read path for fixed-shape ArrayXD columns with nulls

[ebarkhordar]

What does this PR do?

Fixes #8362.

A fixed-shape ArrayXD column (Array2D(shape=(1, 2), dtype="int64"), and so on) that contains a None row is silently cast to float on the python read path: to_dict() / to_pylist() return every value as a float, and integers above 2**53 are altered, because an unrelated row is null. dataset[i][column] returns the correct integers, so the result depends on whether another row happens to be null.

Root cause

ArrayExtensionArray.to_numpy casts a fixed-shape column to float64 so it can insert np.nan for the null rows (a dense numpy array cannot hold both integers and a missing marker). to_pylist returns numpy_arr.tolist(), so that float cast reaches the python read path, where None is available and no cast is needed.

Fix

In to_pylist, when the array is fixed-shape and has nulls, build the list from the nested-list storage (self.storage.to_pylist()), which already carries the exact values and None for the null rows.

The invariant: a non-null value read through the python path must never be altered because another row in the column is null.

Scope kept deliberately narrow:

• The numpy format path is unchanged. to_numpy still returns the dense float64 + np.nan array (asserted by the existing test_array_xd_with_none), so with_format("numpy") behaves exactly as before.
• The dynamic-first-dim branch (shape=(None, ...)) is untouched.

Verification

• New test test_array_xd_with_none_python_format_keeps_integers: fails on main ([[nan, nan]] != None), passes here. It covers integer preservation, the 2**53 + 1 precision case, and asserts the numpy format still returns a dense float64 + nan array.
• python -m pytest tests/features/test_array_xd.py: 103 passed.
• ruff check and ruff format --check clean on the changed files.
• Verified in a clean python:3.11-slim container at HEAD 521a590. Not verified: the torch/tf/jax formatting paths, which this change does not touch (it only affects the python to_pylist path).