Trigger Cs Make Good

Rooster's make-good trigger tool (master spec §3.5: "Agent job: converse, resolve, trigger, escalate"; see app/modules/rooster/make_good_trigger.py): fires this customer's tier-1 make-good event for the current billing month (deterministic, capped one per customer per billing month by make_good_events' own UNIQUE index -- never a second cap invented here) and drafts the apology it promises, unsent, for a human to review and release exactly like any other agent draft (migration 0629's releaser floor covers it the same way).

Gated on the SAME rooster:cs:draft grant as every other CS-drafting route -- deciding a service breach deserves an apology is part of the same drafting capability, not a new one, and is NOT one of the four Constitution-gated launch acts (customer_service/agent_acts.py) -- those are a human (Frederick) flipping a switch per act; this is baseline conversational judgement, the same posture escalate/triage already have.