[core] Support Deletion Vectors for Chain Table

[juntaozhang]

Purpose

#8573

Tests

ChainSplitTest

• testChainSplitSerdeWithDeletionFiles: verifies that ChainSplit can carry deletion files alongside data files and correctly serialize/deserialize them.

SplitSerializerTest

• testV1Compatibility: ensures splits written before this change (without deletion files in ChainSplit) can still be deserialized.
• Updates golden files from v1 to v2 and asserts that ChainSplit round-trips its deletion files, branch mappings, and bucket paths.

ChainTableFileStoreTableTest

• testChainTableWithDelete: batch chain read with DELETE records, covering both deletion-vectors.enabled=true and false.
• testChainTableWithUpdate: batch chain read with UPDATE records and DV enabled.
• testChainTableCrossPartitionDeletionVectors: chain read with group partitions (chain-table.chain-partition-keys) and DV, verifying that deletion vectors are scoped per group.
• testChainTableWithPartialUpdateDelete: confirms that PARTIAL_UPDATE + partial-update.remove-record-on-delete=true still works for chain tables without DV.
• testChainTableRejectsPartialUpdateWithDeletionVectors: validates that enabling DV on a PARTIAL_UPDATE chain table is rejected.
• testChainTableRejectsAggregate / testChainTableRejectsFirstRow: validates that AGGREGATE and FIRST_ROW merge engines are rejected for chain tables.

FlinkChainTableITCase

• testChainTableWithDeletionVectors: end-to-end Flink batch SQL with DV enabled, including snapshot/delta deletes, updates, and compact_chain_table.
• testStreamingReadWithDeletionVectors: Flink streaming read over a chain table with DV, covering the full-load phase (snapshot DV applied) and incremental phase (delta
  cross-partition delete/update changelog).

SparkChainTableITCase

• testChainTableWithDeletionVectors: Spark batch SQL with group partitions and DV enabled, verifying cross-branch deletes and compact_chain_table.