feat: harden missionRuns state machine, heartbeats, retries, and run health dashboard

[krusty-agent]

Summary

• harden missionRuns schema with required P0-6 fields:
  • attempt, parentRunId, durationMs, terminalReason, lastHeartbeatAt, artifactRefs[]
  • plus runtime SLO helpers: heartbeat interval/miss counters + escalation metadata
• implement mission run lifecycle in missionControlCore:
  • create runs, strict state transitions, heartbeat updates
  • heartbeat timeout monitor (degraded after threshold misses, failed after failure threshold)
  • retry baseline with max auto-retry cap and escalation handoff
  • artifact append API with typed refs (screenshot|log|diff|file|url)
• add run visibility queries for dashboards:
  • run listing/filtering
  • aggregate run health metrics (success/intervention/timeout rates + degraded run visibility)
• wire REST v1 endpoints in missionControlApi + router:
  • GET/POST /api/v1/runs
  • POST /api/v1/runs/:runId/{heartbeat|transition|retry|artifacts|pause|kill|escalate|reassign}
  • POST /api/v1/runs/monitor
  • GET /api/v1/dashboard/runs
• expand Team Dashboard runtime operations:
  • add live run list + operator controls wired to /api/v1/runs/:runId/{pause|kill|escalate|reassign}
  • preserve existing team cards/tree and run-health surfaces
• upgrade readiness drill script:
  • scripts/mission-control-readiness-drill.mjs now exercises live run creation + runtime control endpoints (with dry-run fallback)
• expand API key scopes for run controls/visibility:
  • runs:read, runs:write, runs:control, dashboard:read
• update API.md with mission run + dashboard endpoint docs

Mission Control checklist (running)

• P0-1 Foundations (assignees/activity/presence) — merged on main (P0-1 backend: assignees + activity + presence foundation #144/P0-2 frontend: assignee, activity panel, and presence indicators #145)
• P0-2 API keys + agent REST foundations — merged on main (feat(mission-control): scoped API key lifecycle + REST v1 tasks/activity/memory/agents #146)
• P0-3 Memory system — merged on main (Mission Control P2-1: memory schema, APIs, and browser UI #149)
• P0-4 Schedule/calendar scaffolding — merged on main (feat: phase 4 schedule entries + calendar schedule scaffolding #148)
• P0-5 Launch controls + audit events — included in this branch (63c58cb)
• P0-6 Mission runs hardening + run health surfaces — included in this branch (42076f3)
• Runtime control endpoints (pause|kill|escalate|reassign) wired live
• Team Dashboard actions wired to runtime control endpoints
• Readiness drill script now validates real runtime-control responses

Validation

• git fetch origin --prune && git rebase origin/main ✅ (already up to date)
• npx eslint convex/schema.ts convex/missionControlCore.ts convex/missionControlApi.ts convex/http.ts API.md src/pages/TeamDashboard.tsx scripts/mission-control-readiness-drill.mjs ⚠️ expected existing mission-control @typescript-eslint/no-explicit-any baseline remains
• npx tsc -p convex/tsconfig.json --noEmit ⚠️ fails due pre-existing generated API typing drift (didResources missing in generated convex API typings), not introduced by this PR

Notes

This keeps P0-6 missionRuns hardening as one cohesive backend/API/dashboard visibility slice, scoped for merge readiness.