Will writing to a Delta table from a Spark structured streaming job create a version for every micro batch of data written?