A Systematic Investigation of Commonsense Knowledge in Large Language Models