Select and Distill: Selective Dual-Teacher Knowledge Transfer for Continual Learning on Vision-Language Models — arXiv2