[python] Fix Daft Paimon write column alignment

[QuakeWang]

What

• Align Daft write batches to the Paimon target schema by column name before casting.
• Reject missing, extra, or duplicate fields instead of silently writing wrong data.
• Add regression coverage with native pypaimon reader verification.

Tests

• DAFT_RUNNER=native /tmp/paimon-daft-pr6951-venv/bin/pytest -q paimon-python/pypaimon/tests/daft/daft_sink_test.py

Related: Eventual-Inc/Daft#6951

[QuakeWang]

PTAL @chenghuichen

[chenghuichen]

+1.
This is a PR migrated from Daft, and the content is solid — I've already reviewed it there. Daft is order-agnostic on DataFrame columns, while pypaimon requires strict schema match including column order, so aligning by name in the sink is the necessary fix.