fix(ci): flaky subprocess timeouts in unit tests (linux/windows) #163

Several subprocess-spawning unit tests carry timeout budgets tighter than the CI-load cold-start headroom already established for this harness (see the 2026-09-08 comment in test/lib/cli-process.ts bumping the default subprocess timeout 30s->60s for the same class of issue).

Observed failing on PR #160 / #161 CI runs on 2026-09-09, missing their bound by a few hundred ms — consistent with subprocess cold-start variance under parallel CI load, not an actual hang/regression:

• run-process.test.ts: "exits nonzero promptly when the model is unknown" (25s bound, observed 25383ms)
• run-process.test.ts: "unknown stream finish preserves partial output and continues" (30s bound, killed before completion)
• lifecycle.test.ts: "stdin EOF exits cleanly" (5s bound, observed 5300ms on Windows)

Fixed in #162 by widening each timeout with headroom rather than shaving the margin to the exact miss.