To Believe or Not to Believe Y our LLM: Iterative Prompting for Estimating Epistemic Uncertainty