feat: add Loop QA self-testing framework

Structured self-test framework that probes 6 capabilities (tool use,
multistep planning, memory read/write, self-coding, lightning econ) in
round-robin. Reuses existing infra: event_log for persistence,
create_task() for upgrade proposals, capture_error() for crash handling,
and in-memory circuit breaker for failure tracking.

- src/timmy/loop_qa.py: Capability enum, 6 async probes, orchestrator
- src/dashboard/routes/loop_qa.py: JSON + HTMX health endpoints
- HTMX partial polls every 30s on the health panel
- Background scheduler in app.py lifespan
- 25 tests covering probes, orchestrator, health snapshot, routes

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/config.py

@@ -207,6 +207,13 @@ class Settings(BaseSettings):
     thinking_enabled: bool = True
     thinking_interval_seconds: int = 300  # 5 minutes between thoughts
 
+    # ── Loop QA (Self-Testing) ─────────────────────────────────────────
+    # Self-test orchestrator that probes capabilities alongside the thinking loop.
+    loop_qa_enabled: bool = True
+    loop_qa_interval_ticks: int = 5  # run 1 self-test every Nth thinking tick (~25 min)
+    loop_qa_upgrade_threshold: int = 3  # consecutive failures → file task
+    loop_qa_max_per_hour: int = 12  # safety throttle
+
     # ── Paperclip AI — orchestration bridge ────────────────────────────
     # URL where the Paperclip server listens.
     # For VPS deployment behind nginx, use the public domain.