To optimize a join operation in Databricks by ensuring the smaller DataFrame is sent to all executor nodes in the cluster, which function should a data engineer use to mark the DataFrame as small enough to fit in memory on all executors?