convert{su}tof and convertfto{su} tests with float16 need Float16 capability