AVX: vandnpd, vandnps, vblendvpd and vblendvps

[p32blo]

_mm256_andnot_pd also generates a single-precision instruction, so I added the intrinsic to #71

[p32blo]

So, apparently this intrinsics are already implemented at #90, sorry for that.

I will take this opportunity to ask about masks types. Here and here the masks use the __mm* types and I was assuming that, for example, this:

fn blendvpd256(a: f64x4, b: f64x4, mask: f64x4) -> f64x4;

should not be written as this:

fn blendvpd256(a: f64x4, b: f64x4, mask: __m256d) -> f64x4;

Is there a reason for the __mm* types in mask?

And again, sorry for not checking the previous PR

[nominolo]

Regarding mask types, that's also mentioned in #6

[alexcrichton]

Thanks for the PR @p32blo! As you've noticed though I think these were added in #90 as well so I'm going to close this in favor of that.

Let's continue discussion of the mask types on #6 I think?