PeFoMed: Parameter Efficient Fine-tuning on Multimodal Large Language Models for Medical Visual Question Answering