lib: fix AbortSignal.any() observed-composite leak

[bitpshr]

Fixes a memory leak in AbortSignal.any() where an observed composite that follows a long-lived source keeps accumulating in that source's kDependantSignals, even across a forced global.gc().

The retainer is gcPersistentSignals. When you add an abort listener, the observed composite is added there so it stays alive long enough to fire. When one of its sources aborts, abortSignal() marks the composite aborted and runs its abort steps, but it never removes the composite from gcPersistentSignals. That set is only pruned when the listener is explicitly removed or a source is GC'd, so with a long-lived source the aborted composite is retained forever, never collected, and its WeakRef is never pruned from the source's dependant set.

An aborted composite can never fire again, so this drops each transitively-aborted dependent from gcPersistentSignals in the abort path, mirroring the removal that already happens for the directly-aborted signal a few lines down. Once it's collectable, the existing FinalizationRegistry prunes its entry from every source it followed.

The added test observes a long-lived source across many aborted composites and asserts the dependant set drains to zero. It fails before this change (the set stays at the iteration count) and passes after.

Fixes: #64476

[bitpshr]

Hi @geeksilva97 @atlowChemi, wanted to keep this on your radar since you reviewed #62367. This is a direct follow-up: transitively-aborted observed composites stay pinned in gcPersistentSignals, so they never get pruned from a long-lived source's kDependantSignals. Small fix plus a regression test. Whenever you get a chance.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/75046/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.14%. Comparing base (4a5eb1c) to head (157c19a).
⚠️ Report is 9 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #64481      +/-   ##
==========================================
- Coverage   90.15%   90.14%   -0.01%     
==========================================
  Files         744      744              
  Lines      242517   242524       +7     
  Branches    45688    45688              
==========================================
- Hits       218642   218635       -7     
- Misses      15358    15365       +7     
- Partials     8517     8524       +7     

Files with missing lines	Coverage Δ
lib/internal/abort_controller.js	95.34% <100.00%> (+0.05%)	⬆️

... and 25 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.

[geeksilva97]

Thanks, @bitpshr

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/75064/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/75159/

[atlowChemi]

@bitpshr could you try to rebase? I think the Jenkins failure would be fixed by a rebase

[bitpshr]

Rebased onto main, thanks @atlowChemi.

[bitpshr]

Rebased, but the two red checks are unrelated flakes rather than anything from this change. test-macOS failed in the "unusual chars folder" re-run step (a debugger test), and the aarch64 shared-boringssl build hit ERR_INSPECTOR_COMMAND: Promise was collected in test-repl-context (the log shows failed 3 out of 10, so it's flaky). This PR only touches AbortSignal.any() cleanup, so a re-run should clear both.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/75214/