Support clear mm and encoder cache (#33452)

Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>

vllm/engine/protocol.py

@@ -113,6 +113,11 @@ class EngineClient(ABC):
         """Reset the multi-modal cache"""
         ...
 
+    @abstractmethod
+    async def reset_encoder_cache(self) -> None:
+        """Reset the encoder cache"""
+        ...
+
     @abstractmethod
     async def reset_prefix_cache(
         self, reset_running_requests: bool = False, reset_connector: bool = False