Enhancing Knowledge Distillation for LLMs with Response-Priming Prompting