test: fix flaky HTTP/2 reset callback check

[trivikr]

Refs: https://github.com/nodejs/reliability/issues?q=sort%3Aupdated-desc%20test-http2-reset-happy-close

The test previously expected all pending write callbacks to run within two
setImmediate() turns after the HTTP/2 stream emitted 'close'. Write
callbacks can run later when data has already been handed to the socket,
making that timing assumption unreliable under CI load.

Track stream closure and write callback completion independently, and close
the server only after both conditions are satisfied. Use common.mustCall()
to continue verifying that all 16 callbacks run. If a callback is genuinely
lost, the test harness will time out or report the callback count mismatch.

---

Assisted-by: codex:gpt-5.6-sol

[trivikr]

Stress tests on

• main: https://ci.nodejs.org/job/node-stress-single-test/811/
• PR: https://ci.nodejs.org/job/node-stress-single-test/809/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.14%. Comparing base (9024119) to head (b8b5d04).
⚠️ Report is 13 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #64730      +/-   ##
==========================================
- Coverage   90.15%   90.14%   -0.01%     
==========================================
  Files         743      743              
  Lines      242407   242407              
  Branches    45645    45659      +14     
==========================================
- Hits       218532   218527       -5     
- Misses      15357    15384      +27     
+ Partials     8518     8496      -22     

see 35 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.

[inoway46]

LGTM. I specifically checked the removal of the explicit writeCbCount assertion: common.mustCall(..., N_WRITES) and the maybeCloseServer() gate preserve the exact-count invariant, so this removes the timing assumption without weakening the test coverage.