fix(ext/node): improve perf_hooks timerify and add missing exports

[bartlomieju]

Summary

• Rewrites performance.timerify to match Node.js behavior: constructor
  support via new.target, correct .name ('timerified <name>') and
  .length preservation, proper ERR_INVALID_ARG_TYPE validation, and
  optional histogram option
• Exports timerify and eventLoopUtilization as named exports from the
  perf_hooks module, matching the Node.js API surface

Fixes 4 node_compat tests (14 -> 10 failing in the perf_hooks suite):

• test-perf-hooks-timerify-return-value
• test-perf-hooks-timerify-multiple-wrapping
• test-perf-hooks-timerify-invalid-args
• test-perf-hooks-timerify-error

Test plan

• ./x test-compat parallel::test-perf-hooks-timerify-return-value passes
• ./x test-compat parallel::test-perf-hooks-timerify-multiple-wrapping passes
• ./x test-compat parallel::test-perf-hooks-timerify-invalid-args passes
• ./x test-compat parallel::test-perf-hooks-timerify-error passes
• Full parallel::test-perf-hooks suite: 10/14 failing (was 14/14)

[lunadogbot]

Verified the four enabled tests pass against this diff: timerified.name === 'timerified ' + fn.name (chained as timerified timerified m), length preservation, new.target routing for class constructors, and ERR_INVALID_ARG_TYPE for non-function fn and non-record histogram. Two questions inline -- one about the silent loss of the measure entry that the previous impl emitted, one about a potential TypeError: Cannot read properties of null when options is non-undefined-but-null. Neither blocks the upstream test fix.

[bartlomieju]

Thanks for the review!

Re: null options -- Good catch. Added a guard so timerify(fn, null) now throws ERR_INVALID_ARG_TYPE for "options". Pushed in 787795c.

Re: loss of measure entry -- The old performance.measure() call was already non-standard: Node.js emits entries with entryType: 'function', not 'measure'. So the previous behavior was incorrect (it created measure-type entries that no real Node.js code would expect). Removing it is the right thing; proper 'function'-type entry emission is tracked in the TODO for a follow-up.

[fibibot]

The API-shape improvements all match Node's lib/internal/perf/timerify.js:

• validateFunction(fn, 'fn') → equivalent to the new ERR_INVALID_ARG_TYPE throw ✓
• validateObject(options, 'options') → equivalent ✓
• options.histogram validation with the RecordableHistogram arg-type message ✓
• new.target constructor handling (Node uses ReflectConstruct(fn, args, fn), you use new fn(...args) — same observable behavior) ✓
• name = 'timerified ${fn.name}' and length = fn.length ✓
• The four target tests are about exactly this API shape, so they'll pass.

But there's a real semantic regression vs the old code that's worth calling out — see inline.

CI is red on dprint formatting (the long if (options !== undefined ...) line wants to wrap onto multiple lines). ./tools/format.js will fix it; full failure at https://github.com/denoland/deno/actions/runs/25001671896/job/73213243736.

[fibibot]

CI lint fails here — dprint wants this if condition wrapped:

if (
  options !== undefined && (typeof options !== "object" || options === null)
) {
  throw new ERR_INVALID_ARG_TYPE("options", "Object", options);
}

./tools/format.js will auto-fix. Full output at https://github.com/denoland/deno/actions/runs/25001671896/job/73213243736 (3 platforms all show the same one-file diff).

[bartlomieju]

Fixed the formatting -- pushed in 319d669.

Re: the behavioral regression -- agreed this is a defensible tradeoff. The old 'measure' entries were wrong-typed vs Node (which uses 'function' entryType), so anyone relying on them was depending on non-standard behavior that would break when we add proper Node-fidelity 'function' entries anyway.

I'd prefer option 1 (lean into Node fidelity) as a follow-up PR rather than complicating this one -- proper 'function' entry emission needs PerformanceObserver integration with the right entry type, constructor arg tracking, and histogram recording, which is the scope of the tier 2 timerify work. The TODO is tagged to @bartlomieju for tracking.