SIMD vertex buffer decoding and filtering

[yzsolt]

The C++ implementation has hand-rolled x86 (SSSE3, experimental AVX512), ARM (NEON) and WebAssembly SIMD accelerated vertex buffer decoding and filtering.

As of Rust 1.27 there's standardized support for x86 SIMD programming, but NEON and WASM needs external crates and/or nightly rustc.