Beyond Audio: Enhancing SoccerNet-Echoes with Multimodal Event Extraction Using LLMs