deps: add simdutf dependency

[anonrig]

simdutf provides a faster way of providing utf8 operations with SIMD instructions. @nodejs/undici team was looking for a way to validate utf8 input, and this dependency can make it happen.

Edit: I'm proposing either exposing the following functionality through a new module (like node:encoding) or through util.types or buffer

• validate_ascii(string)
• validate_utf8(string)
• count_utf8(string)

PS: simdutf supports more features, and depending on the need, it makes more sense to expose them through a new module, instead of util.types or buffer.

[nodejs-github-bot]

Review requested:

• @nodejs/gyp

[KhafraDev]

This would help speedup both ws and undici's WebSocket implementation (which is still WIP). When we receive a text frame or receive a close frame with a reason, we need to validate that the buffer contains valid utf-8.

There are a few ways of doing so currently: a js implementation by default in both undici and ws, and optionally a package such as utf-8-validate. Note that simdutf is many times faster than the c++ version of utf-8-validate in the benchmark above, and the js fallback version is the slowest.

Here is a PR from @lpinca that shows massive speedups when using simdutf: websockets/utf-8-validate#101. Considering how widespread usage of ws is, exposing a very fast ability to validate utf-8 would improve a ton of the ecosystem.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/48372/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/48377/