Wasm32 support

[yzsolt]

Adds support for WebAssembly SIMD, using the currently available stable intrinsics in Rust.

I also added a GitHub action for automated testing. Benchmarking also works, and shows a ~3x improvement of SIMD vs. scalar when running in wasmtime.

Note that I'm not a SIMD expert, so I may have missed some optimization opportunities when implementing operations which don't map 1:1 to WASM SIMD intrinsics. I've tried to follow already existing SSE and NEON implementations though.

Closes #20.

[arduano]

Request my review when you think it's ready, also please mention how to run/test it

[yzsolt]

Hey, I think it's ready! Rebased onto master and reenabled the commented out scalar tests.

About testing: I've set up things in a way that if you specify wasm32-wasip1 as target, wasmtime (de-facto standard Rust WebAssembly runtime) will be used for running the produced .wasm modules. It's installed on CI automatically, locally you have to do it yourself beforehand. Both cargo test and cargo bench work for wasm32-wasi.

For more complex testing you can build simdnoise for WebAssembly, which I did back when I initially implemented this. It worked nicely, got the same 2-3x scalar -> SIMD speedups in WASM as in native code.

[arduano]

Looks about right, I can't find any issues. Is it good to merge?

[yzsolt]

Looks about right, I can't find any issues. Is it good to merge?

Yes, ship it!