feat(api,ai): #72 OpenAPI 3.1 enrichie + fiabilisation de l'outil AI UI

API (#72):
- backend/openapi_docs.py: 18 tags documentes, assignation auto par prefixe,
  securite bearerAuth/cookieAuth, erreurs 401/403/404/422/500, exemples,
  page /api autonome (liens /docs, /redoc, /openapi.json)
- backend/schemas.py: response_model Pydantic pour ~35 endpoints sans modele
- main.py: FastAPI enrichi + override app.openapi; routes /api et /api/
- ai_routes/bookslm_routes: response_model + doc SSE
- frontend: entree 'API' du menu (i18n FR/EN)
- tests/test_openapi.py (58 tests)

AI UI:
- BooksLM: requetes authentifiees (AuthManager), parsing SSE {token}/error,
  message utilisateur correct (plus le placeholder vide), barre de contexte
- AI Editor: Ctrl/Cmd+J lie une seule fois, libelles i18n, modale de
  reecriture accessible a la place de window.prompt()
- tests/frontend/ai.test.mjs (7 tests) + CI

backend/ai_routes.py

@@ -25,12 +25,13 @@ from backend.ai import (
     ai_summarize,
     ai_translate,
 )
+from backend.schemas import AIStatusResponse
 
 logger = logging.getLogger("obsigate.ai_routes")
 router = APIRouter(prefix="/api/ai", tags=["AI"])
 
 
-@router.get("/status")
+@router.get("/status", response_model=AIStatusResponse)
 async def api_status():
     """Check if AI is configured and which providers are available.