Fix/spark iceberg sort merge

[fkzhao]

Purpose

Fix Spark reads for Fluss lake-enabled primary-key tables.

Brief change log

• Handle empty log-record batches without throwing NoSuchElementException.
• Add a sorted Iceberg reader for lake snapshot and log merge reads.
• Support comparator-only readers when no Iceberg split is available.
• Correct primary-key projection positions during sort-merge reads.
• Preserve timestamp-with-time-zone comparisons.
• Avoid reusing mutable Iceberg row objects during sorting.

Tests

• IcebergLakeSourceTest: 5 tests passed.
• Added regression coverage for:
  • Empty log records.
  • Missing Iceberg split.
  • Projected primary-key rows.
  • Timestamp-with-time-zone primary keys.

API and Format

• No external API changes.
• No data format or storage format changes.
• No migration is required.

Documentation

No documentation changes are required.