Can LLM Watermarks Robustly Prevent Unauthorized Knowledge Distillation?