Core: Select inspect.partitions snapshot metadata by commit time

[fallintoplace]

Rationale for this change

InspectTable.partitions() aggregates files by partition and records the snapshot that most recently updated each partition. Its update condition compared last_updated_snapshot_id with the next snapshot's timestamp_ms. Snapshot IDs are identifiers, not commit-order values, so comparing these unrelated fields made the reported metadata depend on manifest-entry order.

This change follows Apache Iceberg Java's PartitionsTable behavior: compare the current last_updated_at with the snapshot commit timestamp, then update spec_id, last_updated_at, and last_updated_snapshot_id together when the incoming snapshot is newer.

Are these changes tested?

Yes. A parametrized regression test processes the same snapshots in both possible orders. The snapshot IDs deliberately decrease while commit timestamps increase, verifying that commit time alone selects the latest spec_id, timestamp, and snapshot ID.

The following checks pass locally:

• PYTHONPATH=. uv run pytest tests/table/test_inspect.py -q (6 passed)
• PYTHONPATH=. uv run pytest tests/io/test_pyarrow.py -k "inspect_partition" -q (2 passed)
• PYTHONPATH=. uv run --extra datafusion --extra pyiceberg-core pytest tests/table -q (329 passed)
• make lint

Are there any user-facing changes?

Yes. table.inspect.partitions() now reports the latest spec_id, last_updated_at, and last_updated_snapshot_id for each partition independently of manifest-entry order. There are no API changes.