http2: keep reading during small client writes

[mcollina]

Fixes: #54874

[nodejs-github-bot]

Review requested:

• @nodejs/http2
• @nodejs/net
• @nodejs/performance

[codecov]

Codecov Report

❌ Patch coverage is 90.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 89.64%. Comparing base (58a8e1d) to head (4dd0542).
⚠️ Report is 1311 commits behind head on main.

Files with missing lines	Patch %	Lines
src/node_http2.cc	90.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #63009      +/-   ##
==========================================
- Coverage   89.69%   89.64%   -0.05%     
==========================================
  Files         706      708       +2     
  Lines      218222   220331    +2109     
  Branches    41768    42262     +494     
==========================================
+ Hits       195731   197520    +1789     
- Misses      14411    14657     +246     
- Partials     8080     8154      +74     

Files with missing lines	Coverage Δ
src/node_http2.h	91.61% <ø> (ø)
src/node_http2.cc	82.12% <90.00%> (+0.02%)	⬆️

... and 88 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.

[pimterry]

@mcollina I've checked out the branch and the benchmark added here doesn't seem to show any real difference for me with this change. What results are you seeing? Is it that script, or some other benchmark?

[mcollina]

That's the point. I added the bench to verify this did not add a regression.

[pimterry]

This is supposed to be a performance improvement, right? The original issue is about downloads being slower under HTTP/2 than HTTP/1, I assumed the benchmark added was aiming to show that this really fixed the issue. If the one here is just guarding against regressions, can we add another here to test that this fix actually improves things to match HTTP/1?

[github-actions]

This pull request has been marked as stale due to 90 days of inactivity.
It will be automatically closed in 30 days if no further activity occurs. If this is still relevant, please leave a comment or update it to keep it open.

[pimterry]

This is superseded by #65440 which is just landing now - rather than optimizing the specific worst cases, that drops this logic entirely.