test: add mixed-format drop column schema evolution coverage

[liujiwen-up]

Purpose

Linked issue: #258

This PR adds focused coverage for mixed-format schema evolution after DROP COLUMN.

The fixture writes old-schema data files in Parquet and ORC with a dropped column, then drops that column and writes new-schema data files in Avro. The read test verifies that Rust can scan all three formats together while exposing only the remaining table columns.

Brief change log

• Add mixed_format_schema_evolution_drop_column to the Spark provisioning script.
• Write old-schema Paimon data files in Parquet and ORC formats with (id, name, score).
• Drop score, switch to Avro, and write new-schema files with (id, name).
• Add a Rust integration test that:
  • verifies the scan plan includes parquet, orc, and avro data files;
  • reads the table through FileSystemCatalog;
  • checks that the dropped score column is not present in output batches;
  • validates decoded id and name rows across all file formats.

Tests

• cargo fmt --all
• cargo test -p paimon-integration-tests test_read_mixed_format_schema_evolution_drop_column --no-run
• cargo check -p paimon-integration-tests
• python3 -m py_compile dev/spark/provision.py
• git diff --check -- dev/spark/provision.py crates/integration_tests/tests/read_tables.rs

Full runtime execution of the new integration test requires regenerating the Spark-provisioned warehouse with make docker-up.

API and Format

No API or persistent format changes.

Documentation

No documentation changes.