Probing Potential Output : Monetary Policy , Credibility , and Optimal Learning under Uncertainty