Add ChaCha AVX2 implementation

[noloader]

This ticket will track the cut-in of ChaCha AVX2 implementation. There a little more to this cut-in because AVX2 is new territory. I think we have most of the issues sorted out, including old Linux and Solaris with down-level compilers and linkers.

At the moment AVX2 is failing on one case. The case is VS2017, x64, Release build. AppVeyor result at build 20159656. It may be a similar problem as Issue 649.

[noloader]

Cleared over several commits.