Cleanup hash and encryption functions

[Nielsbishere]

Currently the implementation is a bit split because one uses intrinsics and the other doesn't. Even though for some functions this makes sense, it would be easier to generalize them in I32x4 to allow a unified architecture there and make that fallback instead.

[Nielsbishere]

Decision:

• SHA256 should only use the _mm_sha256 instruction. Remove the following:
• 1. [x] _mm_alignr_epi8
• 2. [x] _mm_blend_epi16
• 3. [X] _mm_set_epi64x
• 4. [x] _mm_shuffle_epi8
• [x] GHASH, CRC32 and SHA are okay as a decently optimized version is already presented. Changing this to emulating the instructions exactly will make them slower and will bloat the vector class.
• [x] Move very specific functions such as aes key assist and enc to AES instead of making it accessible for everything. Otherwise it will bloat vector.