Pick an image file and upload it to
/atmosphere/agent/multimodal. The sample
MultiModalAgent (an @Agent class) decodes the
payload, wraps it in a Content.Image, and streams both a binary
frame and a text acknowledgement back over WebSocket.