[python][ray] Rebase exact-range self-merge updates

[XiaoHongbo-Hope]

Purpose

Fix self-merge retries when a concurrent compaction replaces an anchor file but preserves the same (partition, bucket, first_row_id, row_count) range. The old fast path found no files requiring physical rewrite and returned without advancing check_from_snapshot, so the subsequent commit still failed against the compacted-away anchor.

Changes

• Advance staged update messages to the latest checked snapshot even when no file contents need rewriting.
• Run the existing conditional VARIANT migration regression with content_key as an actual partition key.
• Extend the test compaction helper to preserve row-id ranges per partition and bucket.

Tests

• python -m unittest pypaimon.tests.ray_row_id_conflict_rewriter_test pypaimon.tests.ray_data_evolution_merge_into_test (140 passed)
• flake8, py_compile, and git diff --check passed.