Integrate perf enhancers as they mature

[TheKindlerofWildfires]

Keeping an eye on the following

When portable SIMD comes to rust core rust-lang/rust#86656
When bend has good single core performance HigherOrderCO/Bend#391

[TheKindlerofWildfires]

Spent some time working through compiling to the PTX target, it's painful at build time and I'm not sure how the kernels can be invoked 'safely'. One would hope if they are written in safe rust and invoked in safe rust the actual unsafe keyword isn't critical but I'm not certain.

Also if this is going to run on the PTX target it's a no_std environment which makes no_std a required feature

[TheKindlerofWildfires]

Cuda integration is in V1 stage, meaning that parallel on CPU is the big milestone left here