libvpx

Add missing calls to emms in the adaptive quantization code

Also avoid using floating-point operations when adaptive quantization is
disabled.

Change-Id: I54936d7afb661df049cdb3ecd246d04ac2a9d8d3