Clip coefficients in x86 itx asm

Hopefully covers everything. Our tests doesn't cover clipping and I don't really have any idea on how to properly generate coefficients that requires clipping while still being 100% spec-compliant (because clipping is only required in some really obscure edge cases that seems very unlikely to occur with real-world bitstreams).

Also some minor optimizations because I'm too lazy to make two MR:s.

See #103 (closed).