Native scans for Iceberg metadata tables #7431

What is the problem the feature request solves?

Queries on Iceberg metadata tables (.partitions, .snapshots, .all_manifests, .entries, .files, .refs) fall back to Spark. Table-maintenance jobs and partition-listing queries read these, and the operators above the scan stay in Spark.

Describe the potential solution

Read the common metadata tables natively, starting with .partitions, .snapshots and .all_manifests. Check how much iceberg-rust's metadata table support already covers.

Additional context

The native Iceberg scan already supports the _file, _pos, _spec_id and _partition metadata columns. This issue is about the metadata tables, which CometScanRule.isIcebergMetadataTable sends to Spark. #2033 made these queries fall back cleanly. #6719 covers data tables whose names end in a metadata-table name.