A note on random coding bounds for classical-quantum channels