I have a data source, which undergoes frequent CRUD operations. I see from most of the tutorials that the data ingestion are one-time activities like load from file, source from blob, etc. How can I make it periodic so that only the delta new records that were created at source, are ingested into the lakehouse?