bl00mbox: faster sample loading
Most improvement happened in mono samples:
hihat.wav, before: 1077723 us
hihat.wav, halfway: 847702 us
hihat.wav, after: 93141 us
The halfway point above (using an int16 uctypes array) is also applicable to stereo samples:
kick.wav, before: 1402662 us
kick.wav, after: 1109889 us
Notably, using a bytearray and reading only half of the frames and setting pairs of bytes one by one from python is significantly slower, which is why this commit uses both a byterray and an int16 uctypes array.