ci: run all linter benchmarks

[overlookmotel]

After #3221 we can re-enable the linter benchmarks which were disabled in #3172, as they won't hog so much CI time any more.

[overlookmotel]

• ci: run all linter benchmarks #3222 👈
• ci: build each benchmark only with deps it needs #3221
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @overlookmotel and the rest of your teammates on Graphite

[codspeed-hq]

CodSpeed Performance Report

Merging #3222 will not alter performance

_{Comparing 05-10-ci_run_all_linter_benchmarks (31a02a5) with main (5e36e0d)}

Summary

✅ 27 untouched benchmarks

🆕 3 new benchmarks

Benchmarks breakdown

	Benchmark	main	05-10-ci_run_all_linter_benchmarks	Change
🆕	linter[RadixUIAdoptionSection.jsx]	N/A	7.3 ms	N/A
🆕	linter[antd.js]	N/A	10.4 s	N/A
🆕	linter[pdf.mjs]	N/A	1.8 s	N/A

[overlookmotel]

I still don't understand why the linter benchmark takes so long to build. The linter itself is an enormous bunch of code, so you'd expect it, but the benchmark also takes ages to build.

--timings output on CI:

[Boshen]

Github action queues the jobs, adding too many linter benchmarks will drag out these jobs. Let me put this on hold.

Let me fix the generic monomorphization problem first, I'm getting there.

[overlookmotel]

OK, no problem. Actually we should probably drop the 5th benchmark (antd.js) anyway as it takes too long to run even all by itself.

My logic was that now that the jobs don't include building the benchmark and run in ~1 minute, they don't gum up the queue for long, so contention is less of a concern.

[Boshen]

Feels like 2 linter benchmarks is enough for me, closing.