vector shuffles.

[piyush-kurur]

One of the most important class of vector instruction is the shuffle of vector component. It is very useful in implementing fast versions of chacha20 and similar primitives. There are generic enough that they can be implemented in the generic instruction level. It would be good to implement them

Things to figure out

• Sensible representation something like the assignment operator
• Restricting to ones that are commonly implemented.
• Handle questions like endianness here as well