Fusion of Domain-Adapted Vision and Language Models for Medical Visual Question Answering