Add vectorized Iceberg v3 row lineage reads

[osipovartem]

Summary

• inherit Iceberg v3 first_row_id values from data-manifest metadata before snapshot filtering
• expose opt-in _row_id and _last_updated_sequence_number columns from DataFusionTable
• resolve physically stored lineage fields by reserved Iceberg field ID and coalesce them with manifest-derived values
• keep the normal scan path unchanged when lineage columns are disabled

Performance

• lineage synthesis uses vectorized Arrow kernels (add, validity masks, and zip)
• no per-row ScalarValue conversion or expression evaluation
• Parquet reads include lineage dependencies only when the caller explicitly enables the columns
• filters on synthesized lineage columns stay above the scan; unrelated filters retain Parquet pushdown

Validation

• cargo +1.95.0 test -p iceberg-rust --lib --offline table::tests:: -- --nocapture
• cargo +1.95.0 test -p datafusion_iceberg --lib --offline -- --nocapture (71 passed)
• cargo +1.95.0 clippy -p iceberg-rust-spec -p iceberg-rust -p datafusion_iceberg --all-targets --offline -- -D warnings

The integration test creates an Iceberg v3 table, writes rows through DataFusion, reads both lineage columns through a projected Parquet scan, and verifies a filter on synthesized _row_id.

This is the read-side foundation for a correct row-level changelog implementation. It does not claim to implement a full changelog scan or row-lineage-preserving copy-on-write rewrites.