Enhancing LLM Watermark Resilience Against Both Scrubbing and Spoofing Attacks