Add incremental changed data file scans

[osipovartem]

Summary

• add an opt-in scan_changed_data_files mode to DataFusionTableConfig
• select changed data-file events by snapshot provenance across an exclusive-start/inclusive-end range
• normalize an EXISTING entry whose original add snapshot is inside the range to an ADDED event
• retain DELETED entries based on their delete snapshot even when their data sequence predates the range
• expose each event as the partition-constant __iceberg_change_file_status column
• inherit missing manifest-entry snapshot IDs according to the Iceberg specification

Performance and correctness

• normal current-state scans keep their existing path
• range membership and status normalization run once per manifest entry, never per row
• only manifests in the requested snapshot range and their relevant data files are read
• change status is a per-file partition constant
• equality and position delete files are not misreported as changed data rows

Validation

• full datafusion_iceberg suite: 72 passed
• strict clippy for iceberg-rust and datafusion_iceberg with -D warnings
• Iceberg v3 integration test covers a range spanning two consecutive append snapshots and verifies all rows, inherited row IDs, update sequence numbers, and ADDED statuses
• Rustice end-to-end CHANGES range test passes against this revision

Rustice uses this read-side primitive to classify Snowflake-compatible CHANGES(INFORMATION => DEFAULT) rows. Row-level classification remains outside this library PR.