A data lake needs supported SQL updates and deletes, transactionally consistent table changes, and snapshot-based table behavior over S3 data. Which option is the most direct fit when using supported Athena functionality?

1. An Apache Iceberg table configured for supported Athena operations.
2. A folder of unrelated CSV objects with no table transaction metadata.
3. Use a normal external Parquet table and independently overwrite data files without any transactional table metadata.
4. S3 Versioning alone, treated as a complete SQL transaction manager for every row.