Copying ScyllaDB data to S3, using Spark, performance optimization