Fix misleading prefill docstring in _llm_runner.pyi

The old wording suggested calling generate("") after prefill, but
the Python API takes List[MultimodalInput] and the C++ runner
enforces a non-empty prompt. Updated to describe the correct
post-prefill pattern.

This PR was authored with the assistance of Claude.

Author: kirklandsign

extension/llm/runner/_llm_runner.pyi

@@ -479,8 +479,8 @@ class MultimodalRunner:
     def prefill(self, inputs: List[MultimodalInput]) -> None:
         """
         Prefill multimodal inputs (e.g., to rebuild KV cache from chat history)
-        without generating tokens. After prefill, call generate("") to start
-        decoding from the prefilled state.
+        without generating tokens. After prefill, call generate() with a
+        non-empty final text input to start decoding.
 
         Args:
             inputs: List of multimodal inputs to prefill