feat(vector): complete data evolution support

[ChaomingZhangCN]

Purpose

Linked issue: apache#197

Implement Phase 2 VECTOR data-evolution support for append-only Parquet tables:

• allow VECTOR fields in data-evolution tables;
• add a validated SchemaManager commit path for add/drop evolution;
• preserve VECTOR reads across historical schema files and partial-column updates;
• reject incompatible VECTOR dimension or element-type changes before persisting the next schema;
• reject VECTOR comparator roles including sequence and sequence-group ordering fields, while retaining primary, partition, and bucket-key validation.

Tests

• Built paimon-common-test, paimon-core-test, paimon-data-evolution-table-test, paimon-write-and-read-inte-test, and paimon-blob-table-inte-test
• Related schema/VECTOR unit suites: 96 passed
• Complete Data Evolution integration suite: 198 passed, 6 expected skips
• Complete write/read integration suite: 144 passed
• Complete BLOB/Data Evolution integration suite: 162 passed
• python3 -m pre_commit run --files <changed-files>
• git diff --check origin/main...HEAD

API and Format

No public API, storage-format, or protocol changes. VECTOR continues using the existing Parquet LIST representation. The new schema-commit entry point is internal to the core schema manager.

Documentation

Updated the VECTOR data-type documentation with validated data-evolution behavior and comparator restrictions.

Generative AI tooling

Generated-by: Cursor Agent GPT-5.6 Sol