To minimize read and write latency when optimizing a Spark job for processing a large dataset stored in Azure Data Lake Storage Gen2, which configuration should be prioritized?