Will it be possible to add bfloat16 and float16 support at some point? Will be very helpful when working with large point clouds. Thanks!