Investigating Large Language Models for Code Vulnerability Detection: An Experimental Study