[vector API] faster sorting using AVX512 permutations

[John Rose]

Here's some fun reading about optimizing sort algorithms
using vectorized local permutations (selected from a schedule
or table of static permutation vectors) to sort data in one
or more vectors, branch-free.

https://arxiv.org/pdf/1704.08579.pdf
https://gitlab.mpcdf.mpg.de/bbramas/avx-512-sort
https://hal.inria.fr/hal-01512970v1/document

— John