Implement native support for Iceberg _file metadata column #2175

Describe
Iceberg scans in Auron currently fall back when the projected schema contains Iceberg metadata columns, even for cases where file-level metadata such as _file could be provided natively.

This prevents queries like select _file from iceberg_table or select data_col, _file from iceberg_table from using the native Iceberg scan path.

Describe the solution you'd like
Add native support for Iceberg metadata columns that can be represented as file-level constant values, starting with _file.

A possible approach is:

• stop treating all Iceberg metadata columns as unsupported by default
• distinguish between:
  • file-backed data columns
  • metadata columns that can be materialized outside the file payload
• extend IcebergScanPlan to carry both the file schema and metadata/extra column schema
• pass supported metadata values through the native file-scan path, for example via per-file partition/constant values
• project both normal data columns and supported metadata columns from the native Iceberg scan
• continue to fall back for unsupported row-level metadata columns such as _pos

Additional context
This should be implemented conservatively.

A good initial scope is:

• support _file
• keep _pos and other row-level metadata columns on the fallback path

Suggested regression coverage:

• select _file from iceberg_table uses NativeIcebergTableScan
• select id, _file from iceberg_table uses NativeIcebergTableScan
• select _pos from iceberg_table still falls back
• native results match vanilla Spark results