More optimizations

Today I've optimised decoding videos with alpha channels by skipping YUV->RGB conversion for pixels whose alpha value is 0, thus improving pixel conversion speed :)

Also, I've switched the YUV->RGBA conversion in AVFoundation backend to use apple's BGRX decoding and then converting that to RGBA instead of using the project's own YUV conversion functions because apple's implementation works better on iOS.

i used kernel byte swap functions on iOS and Mac to optimise this and it performs better on actual devices. Still not perfect, but better. If it were any better, it'd have to be written in assembly (which is a viable option for future versions of this library)