In a Delta Live Tables pipeline, a data engineer has set up a data quality constraint as follows: CONSTRAINT valid_id EXPECT (id IS NOT NULL) _____________. How should the blank be filled to ensure that records violating this constraint are dropped and their violations are reported in metrics?