[core] Add committer-side bucket consistency check

[Aitozi]

Purpose

Add committer-side bucket consistency validation for write-only unordered append tables.

after #6741 When bucket-append-ordered=false and write-only=true, writers skip restoring previous files, so bucket-count validation can be bypassed after bucket rescale. This change adds an internal commit-side checkSameBucket path for fixed hash bucket tables to validate touched partitions before committing.

The check is integrated with ConflictDetection, reuses the existing conflict path when available, and uses a bounded partition cache to avoid repeatedly checking the same partition within one committer lifecycle.

Tests

• Added core coverage for unordered write-only append:
  • prepareCommit succeeds when bucket count changes.
  • commit fails for an existing partition with mismatched bucket count.
  • writing a new partition succeeds.
  • append-style commits with both DELETE and ADD still validate bucket consistency.
• Added Flink IT coverage:
  • INSERT INTO fails after bucket count changes.
  • INSERT OVERWRITE succeeds for rescaling.