node:http: 2 remaining byte-level diffs (timeout-overflow, many-ended-pipelines) — final piece of #2132 original scope

[proggeramlug]

Summary

The last two tests in #2132's original scope — the only remaining diff cases (both runtimes exit 0, stdout bytes disagree). The byte-level header/framing work across #2184/#2197/#2259/#2335/#2416/#2461 closed the rest; these two are what's left of the literal "byte-level output diffs".

test-http-many-ended-pipelines.js
test-http-timeout-overflow.js

Both currently show (no output) from Perry where Node emits content — so they may actually be silent-exit cases mis-tallied as diff (Node stdout non-empty, Perry empty → "differs"). Confirm whether Perry is producing wrong bytes vs. producing nothing; if nothing, they fold into the #4909 silent-exit triage.

• test-http-timeout-overflow.js — exercises setTimeout with an out-of-range (overflow) delay; Node clamps to 1ms and warns. Likely a timer-overflow-clamp behavior gap.
• test-http-many-ended-pipelines.js — many pipelined requests each ended; output/ordering parity.

Acceptance

Diff Perry vs Node byte output for both; either fix the divergence or reclassify into #4909. This is the final piece of #2132's stated (byte-diff) scope.

---

Part of #2132. Measured via scripts/node_core_subset.py --api http https vs pinned Node v22 corpus.