watchDrill(fetch, sleep) polls a forge until it is healthy again and returns each distinct line it said while unhealthy #2393

Part of #2350, which a builder left without a pull.

Done looks like:

• [ ] A pure module scripts/drills/watch.ts exports watchDrill({url, fetch, sleep, limit}). It polls /api/v1/health and / once per tick and keeps each distinct status and text seen while unhealthy. It returns {startedAt, healthyAt, lines}, or a timeout. It has no stdin, no posting and no globals.
• [ ] Test (in GATE) with an injected fetch that goes unhealthy and comes back. It asserts the lines and that the result arrives without a real wait.

Parked by the crew: the check failed ("the pages > (unnamed)") and the decider saw no point in another try (pull #2402 closed). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.

Parked by the crew: the check failed ("token counts on the pages > (unnamed)") and the decider saw no point in another try (pull #2411 closed).

Parked twice since a person last spoke here, so it is the planner's now. Every reason:

1. the check failed ("the pages > (unnamed)") and the decider saw no point in another try (pull #2402 closed)
2. the check failed ("token counts on the pages > (unnamed)") and the decider saw no point in another try (pull #2411 closed)

The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.

replaced by #2424, the same work. Both reds ('the pages', 'token counts on the pages') were unrelated to a standalone script, so it now waits for the runner fix #2416.