Pareto-Optimality, Smoothness, and Stochasticity in Learning-Augmented One-Max-Search