A data engineer is configuring a pipeline that will potentially see late-arriving, duplicate records. In addition to de-duplicating records within the batch, which of the following approaches allows the data engineer to deduplicate data against previously processed records as it is inserted into a Delta table?