John the Ripper SIMD support enhancements

by Lei Zhang for Openwall

In this project, we intend to improve John the Ripper's SIMD support. We want to enhance the use of AVX2 intrinsics, and add support for AVX-512 instructions. As AVX-512 is not available yet, we'll use MIC as our experiment target. We'll also add SIMD support for non-x86 architectures, which should be straightforward since the pattern is similar. Another goal is to improve the SIMD implementation of the SHA family hashes. We may try bitslice and other low-level optimizing techniques.