Optimal Choice under Costly Precision: Entropy Regularization as a Special Case