Decision Making in Non-Stationary Environments with Policy-Augmented Search