Add AI provider degradation registry and alerting (server-26#14)

Three AI dependency failures in one night (retired Gemini model IDs,
depleted Gemini balance, unpayable OpenAI account) each surfaced only
as a single ERROR log line that nobody was watching. Add
app/internal/ai_health.py, a shared in-memory registry that
transcription.py and llm_correlator.py report into on every call
(success and failure), distinguishing permanent conditions (dead
model, dead billing) which alert immediately from transient ones
(rate limits, network blips) which only alert after they persist.
Alerts POST once per degradation episode and once on recovery to an
optional Discord webhook (AI_ALERT_WEBHOOK_URL), reusing alerter.py's
httpx pattern. State is exposed unauthenticated at GET /health/ai
alongside the existing /health.

Closes logan/server-26#14

drb-c2-core/app/main.py

@@ -8,6 +8,7 @@ from app.internal.node_sweeper import sweeper_loop
 from app.internal.summarizer import summarizer_loop
 from app.internal.vocabulary_learner import vocabulary_induction_loop
 from app.internal.recorrelation_sweep import recorrelation_loop
+from app.internal import ai_health
 from app.config import settings
 from app.internal.auth import (
     require_firebase_token,
@@ -120,3 +121,12 @@ app.include_router(media.router)
 @app.get("/health")
 async def health():
     return {"ok": True, "mqtt_connected": mqtt_handler.is_connected}
+
+
+# Deliberately unauthenticated, same as /health above: the CI deploy step
+# curls /health with no credentials, and this is diagnostic state (which AI
+# tier is degraded and why), not a secret — no API keys or tokens appear in
+# it. Keeping it auth-free means an external uptime check can watch it too.
+@app.get("/health/ai")
+async def health_ai():
+    return {"tiers": ai_health.snapshot()}