aarch64 cabac optimizations

Looks good overall; I didn't track each individual change in the final commit though as that's quite a bit of work...

I was assuming nobody is going to change the cabac context memory layout. On the other hand I thought of testing if exchanging low and range makes using LDP a worthwhile change. I'll add compile time assert for this.

I think it's easier to review the new code than the changes. sorry.