Multi-LogiEval: Towards Evaluating Multi-Step Logical Reasoning Ability of Large Language Models