Improve VLLM model discovery error handling

• Add comprehensive error handling in check_model_availability method
• Provide helpful error messages with actionable solutions for 404 errors
• Warn when API token is set but model discovery is disabled

llama_stack/providers/utils/inference/model_registry.py

@@ -100,7 +100,7 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
     async def should_refresh_models(self) -> bool:
         return False
 
-    async def allow_listing_models(self) -> bool:
+    async def enable_model_discovery(self) -> bool:
         return True
 
     def get_provider_model_id(self, identifier: str) -> str | None: