Benchmark

[IWANABETHATGUY]

Benchmark

Old Solution

Firstly, we compile rspack to node binding via napi-rs and then bench our rspack using nodejs benchmark.
We would expose the nodejs api to the user side finally, so using nodejs benchmark framework for performance monitoring makes sense.
But this solution has some downsides,

1. More runtime involves more benchmark noise will introduce, this solution has at least three benchmark noise, rust side(IO, async runtime), node binding(data transform), Node, and V8(GC, JIT), for example:
   run benchmark twice without any code change:
   
   1.5s is insane.
2. We may need to rerun the benchmark multiple times to verify if we introduce performance regression in some commits, also
   as we all know, the benchmark is a time costing task in CI workflow, triggering the benchmark every commit also extends our workflow.

New Solution

We keep the old nodejs benchmark as is. Some changes are listed below:

1. We bench our rspack on the rust side via criterion by default, triggering by hand via GitHub issue comment before you are ready to review (you could trigger any time as you want, usually we trigger it before reviewing).
2. If you modify the node_binding crate in pull request, you should trigger the nodejs benchmark. (You could trigger both the rust benchmark and nodejs benchmark if you want)
3. Setting up the GitHub cron action that trigger every day for the nodejs benchmark.