ISO Loader 0.8.x Beta + preview of DreamShell v4.0 RC 5

[SWAT]

New build: https://github.com/DC-SWAT/DreamShell/re...0.0RC5-T29
Just need to check that nothing is broken and that it worked as before.

- New optimized algorithm for separating PCM channels (2x faster).
- New optimized algorithm for SQ mode with separating PCM channels in one process (also saves half of RAM for CDDA buffer).
- Updated KallistiOS to latest.

Now memory usage for CDDA looks like this:
- No IRQ and AICA DMA: 32KB CPU RAM and 32KB SPU RAM.
- No IRQ and SQ: 16KB CPU RAM and 32KB SPU RAM.
- IRQ and AICA DMA: 16KB CPU RAM and 16KB SPU RAM.
- IRQ and SQ: 8KB CPU RAM and 16KB SPU RAM.

For better perfomance should use DMA for both IDE and AICA SPU RAM. Because SQ as a PIO mode wastes CPU time.
But for some games we need to use SQ. And now SQ mode not only wastes CPU time, but also saves memory.
I came close to adding CDDA support to low-level mode. The new PCM splitting algorithm with SQ can be combined with PIO reading and completely eliminate the buffer in the CPU RAM. This will not be very optimal in terms of performance and will not work everywhere, but the main thing is that it will help games such as Looney Tunes.
But to do this, I really still need to fit the CDDA emulation code into the remaining approx 4 kilobytes in low-level mode. I'll leave this challenge for later )

@CrisMod
Looney Tunes works only from 8c004800 address with loader without CDDA.
For this game no free memory for CDDA emulation at all.
At least for current version of the loader.