# Implement FastAPI Lifespan Model Caching
Loading machine learning models from disk on every HTTP request introduces massive latency. Modern FastAPI applications use lifespan handlers (`@asynccontextmanager`) to load models into an application cache dictionary at startup and clean them up during shutdown.
### Task
Create an async lifespan context manager `create_model_lifespan(model_loader_fn, cache_dict: dict)` that:
1. Calls `model_loader_fn()` and stores the returned model in `cache_dict["model"]`.
2. Yields control to the running application.
3. Clears `cache_dict` when the application shuts down.