test: add mixed-format schema evolution coverage

[QuakeWang]

Purpose

Linked issue: #258

Rust reader already had coverage for Java/Spark-generated full type boundary values, but schema evolution compatibility across mixed data file formats was not explicitly covered. In particular, existing fixtures did not verify that Parquet/ORC/Avro files written under different schema versions are read consistently by field id, with null-fill for added columns and casts for promoted types.

Brief change log

• Add format_schema_evolution_add_column Spark fixture with Parquet old-schema files and ORC/Avro new-schema files.
• Add format_schema_evolution_type_promotion Spark fixture with Parquet INT files and ORC/Avro BIGINT files.
• Add Rust integration tests that assert:
  • planned files include Parquet, ORC, and Avro
  • planned files include multiple schema ids
  • reader output preserves null-fill and INT to BIGINT cast semantics

Tests

• cargo test -p paimon-integration-tests test_read_format_schema_evolution -- --nocapture

API and Format

Documentation