Apply the slow-platform allowance where the bounds are actually written

[its-janghoon]

Follow-up to #47, which was incomplete — and the next Windows run said so.

What #47 missed

It scaled only the default child timeout. The very test that was flaking passes its own timeoutMs: 30_000, so it stayed pinned to thirty seconds and failed again at 30872ms on the run that already contained the fix. That number is the evidence: with the allowance in effect the bound would have been 90s.

Two gaps, both closed by construction

cliIt.concurrent now treats a caller's test timeout as a FLOOR, not a ceiling. Thirty-seven call sites pass 60_000 — comfortably above the old fixed 30s child bound and below the Windows one, so honouring them literally reintroduced exactly the drift the derived constants exist to prevent. One clamp fixes all thirty-seven; editing them individually would not have kept the thirty-eighth honest.

slowPlatform(ms) is exported for the call sites that genuinely need their own bound, so intent stays readable while the correction is applied for them.

The duration assertion beside one of those bounds is scaled with it — its claim is that the CLI exits promptly rather than being killed, and a duration pinned to a base figure while the bound moves tests something else.

Still a mitigation

Worth repeating from #47: the platform factor responds to a timeout, not to a diagnosed cause. What #47 genuinely fixed is that the next occurrence will report what the child printed — both streams were being discarded on the timeout path at two separate layers. If it recurs with output preserved, that output is the diagnosis and this factor should be revisited against it.

Gates

typecheck        0 errors across the monorepo
packages/redrob  3380 pass / 0 fail
run-process      13/13