diagnostics_channel: add threadpool work event

[IlyasShabi]

This PR aims to add support for experimental threadpool.work.<type> diagnostics channel.
type: kind of threadpool work

The event exposes:

• enqueued: submission timestamp
• started: execution start timestamp
• ended: execution end timestamp

These timestamps allow diagnostics tools to distinguish queue time from native execution time. Cancelled work reports started and ended as null.

The channel and subscriber state are cached per Environment so the unsubscribed path remains minimal.

Benchmark

Unsubscribed configurations showed no significant regressions. Most subscribed configurations were also within noise; zlib at n=10000 measured a 3.48% regression on local machine (will share the onces from CI)

[nodejs-github-bot]

Review requested:

• @nodejs/performance

[mcollina]

while this is a good generic API, I'm not convinced. Shouldn't we add one for each usage?

cc @bengl @Qard ?

[IlyasShabi]

@mcollina You mean separate channels named threadpool.work.<type>? This is also a valid approach and would let consumers subscribe only to relevant work and remove type from the payload.

[codecov]

Codecov Report

❌ Patch coverage is 88.60759% with 9 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.40%. Comparing base (fcfb7ec) to head (a1c60b9).

Files with missing lines	Patch %	Lines
src/threadpoolwork-inl.h	82.85%	0 Missing and 6 partials ⚠️
src/env.cc	84.21%	0 Missing and 3 partials ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #65982   +/-   ##
=======================================
  Coverage   90.40%   90.40%           
=======================================
  Files         791      791           
  Lines      276226   276293   +67     
  Branches    53045    53053    +8     
=======================================
+ Hits       249732   249796   +64     
+ Misses      16899    16896    -3     
- Partials     9595     9601    +6     

Files with missing lines	Coverage Δ
src/crypto/crypto_util.h	66.66% <100.00%> (ø)
src/env.h	97.53% <100.00%> (+0.12%)	⬆️
src/glob/node_glob.cc	75.45% <100.00%> (ø)
src/histogram.cc	86.32% <100.00%> (-0.44%)	⬇️
src/node_api.cc	77.07% <100.00%> (ø)
src/node_diagnostics_channel.cc	85.71% <100.00%> (+1.09%)	⬆️
src/node_diagnostics_channel.h	87.50% <100.00%> (+16.07%)	⬆️
src/node_file.cc	75.57% <100.00%> (ø)
src/node_internals.h	82.25% <100.00%> (+1.90%)	⬆️
src/node_sqlite.cc	81.89% <100.00%> (-0.03%)	⬇️
... and 3 more

... and 24 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[mcollina]

yes

[mcollina]

I think you are missing the documentation

[IlyasShabi]

I added some doc to https://github.com/nodejs/node/pull/65982/changes#diff-d9c2d2d884d1dea00c82e7f00176441f9524573b6429fd83bf3835cd09a5b029R2029 could you point me to the missing doc?

[IlyasShabi]

@mcollina could you please have a look?

[bengl]

Make it clear that this is the libuv thread pool. I'd even suggest putting it the event name to make it clearer. While it may be self-evident from reading the supported values, being as specific as possible is helpful in both docs and channel names.