[noise] hkdf branch with arbitrary-length keyed hashing functions

[Jason A. Donenfeld]

On Mon, Oct 12, 2015 at 8:17 PM, Jason A. Donenfeld <Jason at zx2c4.com> wrote:
> Any objections or security considerations with regards to this?

*If* this turns out to be safe, we've also won a huge performance
speedup, making this variant of the hkdf branch actually faster than,
rather than slower than, the n0 branch.

For each MixKey():
  - n0: 1 or 0 encryption + 2 hash
  - hkdf: 6 hash
  - blake2b-kdf: 1 hash

Success.

Now it's that *if* there I'm anxious to hear about.