When working with Delta Lake, which indexing technique is most effective for optimizing join performance on columns characterized by high cardinality?