`FourSimdBlocks` for 16-byte archs

[V0ldek]

Is your feature request related to a problem? Please describe.
On a theoretical 64-bit target with SSE2 support it might be useful to consider four SIMD blocks at a time, since the SSE2 vectors are 16-byte wide, so a combined movemask on four of them gives a single 64-bit int.

Describe the solution you'd like
There is a FourSimdBlocks type that is guaranteed to be four times larger than SimdBlock. It should have an impl of a

pub fn blocks(&self) -> (&AlignedBlock<SimdBlock>, &AlignedBlock<SimdBlock>, &AlignedBlock<SimdBlock>, &AlignedBlock<SimdBlock>)

function, giving all the individual blocks.

It might also be beneficial to have a function that returns two TwoSimdBlock-aligned blocks.