Automatic distribution for non-bucketed tables during streaming ingest

Currently non-bucketed tables write to a single file in transaction batch. This limits the concurrency of compactor job to be 1. Provide an option to automatically distribute the incoming records to N files within a transaction batch.