Visualization Literacy of Multimodal Large Language Models: A Comparative Study