perf(fetch): skip empty ResponseInit work and send buffered bodies directly

[anonrig]

This relates to...

• Follow-up to perf(fetch): skip empty RequestInit work #5701 (same class of “defaults / scaffolding do extra work”)
• GHSA-35p6-xmwp-9g52 (idle-socket validation must still yield to poll)

Rationale

#5701 skipped Request constructor work when RequestInit was empty (dictionary defaults were making every fetch(url) look non-empty) and dropped the idle-socket setTimeout(0) floor.

This PR applies the same ideas to the next similar hot paths, independently of that branch:

1. ResponseInit defaults — status: 200 and statusText: "" meant every new Response() / new Response(body) ran WebIDL conversion and reason-phrase validation for values makeResponse() already has.
2. Buffered request bodies — extractBody() already keeps string / BufferSource bodies on source, but HTTP-network-fetch still teed the companion Web Stream and incrementally read it before handing bytes to the dispatcher.
3. Idle-socket validation start time — validation was only scheduled when the next request arrived. Starting it when the socket becomes idle lets the GHSA yield overlap body consumption instead of gating the next dispatch.

Changes

• Treat “init has members” as user-specified ResponseInit fields, not WebIDL defaults. Omitted / {} init skips conversion; makeResponse() already supplies 200 / "".
• If request.body.source is a string or Uint8Array, clone headers only (no stream tee) and pass source to agent.dispatch.
• Schedule H1 idle-socket validation in onMessageComplete when kSocketUsed becomes true. The existing write-time clearIdleSocketValidation and poisoning checks are unchanged.

Features

N/A

Bug Fixes

N/A

Breaking Changes and Depreciations

N/A

Benchmarks

Loopback, Node 22.14. Measured against main (e609acb0):

	main	this PR
new Response()	4.1M ops/s	11.1M ops/s (~2.7×)
new Response(null, { status: 201 })	3.1M ops/s	2.9M ops/s (unchanged path)
fetch POST string body p50	1.66 ms	1.49 ms
sequential fetch().text() p50	1.48 ms	1.45 ms

The GET keep-alive number is still dominated by setTimeout(0) on main; #5701 replaces that with setImmediate. Starting validation at idle is complementary once that lands.

Status

• I have read and agreed to the Developer's Certificate of Origin
• Tested
• Benchmarked (optional)
• Documented
• Review ready
• In review
• Merge ready

[codecov-commenter]

Codecov Report

❌ Patch coverage is 97.91667% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.47%. Comparing base (f387057) to head (c75d899).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/web/fetch/index.js	92.30%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5702      +/-   ##
==========================================
- Coverage   93.47%   93.47%   -0.01%     
==========================================
  Files         110      110              
  Lines       38846    38929      +83     
==========================================
+ Hits        36310    36387      +77     
- Misses       2536     2542       +6     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 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]

Can you please split the changes to client-h1 from the fetch ones?

[mcollina]

Can you send the changes to this as a separate PR?