MemVL T: Vision-Language Tracking with Adaptive Memory-based Prompts