Which function is appropriate for estimating the approximate number of distinct values in a table containing trillions of rows?