SPIP: Row-level operations in Data Source V2

Row-level operations such as UPDATE, DELETE, MERGE are becoming more and more important for modern Big Data workflows. Use cases include but are not limited to deleting a set of records for regulatory compliance, updating a set of records to fix an issue in the ingestion pipeline, applying changes in a transaction log to a fact table. Row-level operations allow users to easily express their use cases that would otherwise require much more SQL. Common patterns for updating partitions are to read, union, and overwrite or read, diff, and append. Using commands like MERGE, these operations are easier to express and can be more efficient to run.

Hive supports MERGE and Spark should implement similar support.

SPIP: https://docs.google.com/document/d/12Ywmc47j3l2WF4anG5vL4qlrhT2OKigb7_EbIKhxg60