This makes the mapping from 16-bit color to LUT entry straightforward, while still fully defining the entire range from 0x0000 through 0xFFFF. This fixes an edge case where colors greater than 0xFF00 would wrap past the end of the R/G LUTs onto the next color channel.

Almost fast enough.. just 1.3ms more to shave off per frame.

Originally I wasn't planning on doing this, but since I need to reclaim some RAM anyway, might as well just start hacking out everything we don't need...