Returns the smallest period that contains p1 and p2.
This differs from regular period union in a critical ways: It won't throw an error for non-adjacent p1 and p2, but just absorb the intervening values into the result period.