MFC-Bench: Benchmarking Multimodal Fact-Checking with Large Vision-Language Models