Towards Safe AI Clinicians: A Comprehensive Study on Large Language Model Jailbreaking in Healthcare