fix: normalize dictionary types in Arrow scans

[GayathriSrividya]

Rationale

Iceberg treats Arrow dictionary encoding as an encoding detail rather than a separate logical type. However, ArrowScan.to_table currently concatenates batches without decoding dictionary-encoded columns first. A table containing both plain strings and dictionary-encoded strings therefore fails to scan with ArrowTypeError: Unable to merge.

This can occur in production when files written with dictionary encoding are later rewritten by Athena or Trino optimization into plain strings.

Changes

• Recursively unwrap Arrow dictionary types while preserving unrelated Arrow types and schema metadata.
• Normalize dictionary-encoded batches before permissive concatenation in ArrowScan.to_table.
• Add regression coverage for mixed plain/dictionary string batches and nested dictionary types.

Attribution

I checked the issue and PR history before opening this PR. I did not find an earlier PR or implementation to cherry-pick for #3260.

Verification

• make lint
• make test (3711 passed, 1534 deselected)
• uv run python -m pytest tests/io/test_pyarrow.py -q -k "mixed_dictionary or ensure_non_dictionary"