Evaluating Open-Source Vision-Language Models for Multimodal Sarcasm Detection