fix: rename generation model setting to ollama_llm_model to avoid OLLAMA_MODEL env collision

The deployed container sets OLLAMA_MODEL=nomic-embed-text for embeddings,
which shadowed the generation-model setting and broke Phase 0 keyword
extraction and Phase 4 LLM re-ranking on every request. The setting is now
ollama_llm_model (env: OLLAMA_LLM_MODEL, default gemma4:e2b), startup logs
the resolved generation model, and Phase 0/Phase 4 LLM calls are wrapped in
a 12s asyncio.wait_for with graceful fallback so a hung call cannot gate
retrieval for the full 120s client timeout.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/main.py

@@ -142,7 +142,7 @@ async def health(settings: Settings = Depends(get_settings)) -> HealthResponse:
             },
             "ollama": {
                 "url": settings.ollama_url,
-                "model": settings.ollama_model,
+                "model": settings.ollama_llm_model,
                 "healthy": service_health.get("ollama", False)
             }
         }
@@ -412,6 +412,7 @@ async def startup_event():
     logger.info(f"Wiki.js: {settings.wikijs_url}")
     logger.info(f"SearXNG: {settings.searxng_url}")
     logger.info(f"Ollama: {settings.ollama_url}")
+    logger.info(f"Ollama generation model: {settings.ollama_llm_model} (embedding model: {settings.ollama_embedding_model})")
 
     # Initialize all service clients
     await startup_clients()