Behavioral Fingerprinting of Large Language Models