Optimizing FreeSWITCH audio muxing

Surprisingly enough, as much as FreeSwitch is using optimized libraries such as VP9 for video processing, a lot of internal audio processing still relies on relatively unoptimized C code. My work with Jack2 showed that you can squeeze a lot of extra audio muxing in a single core/thread with a little bit of work.

So I've started implementing audio optimization (SIMD operations in x86_64 / AVX / AVX2 processors) for FreeSwitch. The code might or might not make it into mainline, depending on what Signalwire's priorities are, but I intend to eventually make this available under the same license as FreeSwitch.

In the meantime I'm happy to share with people interested in helping by testing this out — you'll need to at least know how to compile FS and do load testing for specific operations in your environment — send me a message if this you!