fix(parquet): fallback nested field reading to RowGroup reading (walkaround)

[zhf999]

Purpose

• Fix a correctness issue when predicate pushdown enables page-level filtering on Parquet files containing nested columns.
• When read schema contains nested fields (struct/list/map), page-level predicate filtering currently uses BuildPageFilteredSchema mapping that does not correctly align nested Parquet leaf columns, causing nested column reads to fail.
• Add a fallback path: when nested columns exist in read schema, disable page-level index filtering and fall back to row-group level reading, which preserves correctness.

Changes

• In src/paimon/format/parquet/parquet_file_batch_reader.cpp:
  • Detect nested columns in file schema via ArrowSchemaValidator::IsNestedType.
  • Skip page-level index filter initialization/build when nested field exists.
  • This makes reader use row-group filtered flow for nested schemas, avoiding the failing nested leaf-index mapping path.

Tests

• Added/updated unit tests in src/paimon/format/parquet/page_filtered_row_group_reader_test.cpp (no public API test harness changes).
• Validated expected behavior on:
  • nested struct full read with predicate
  • nested struct only-read with predicate on non-selected top-level field
  • nested list read with predicate
  • nested map read with predicate
  • adjacent nested columns coexistence

API and Format

• No API changes.
• No storage format/protocol changes.

Documentation

• No user-facing feature/documentation updates required.

Generative AI tooling

• No

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[lxy-9602]

+1