Learning-Augmented Online Bipartite Matching in the Random Arrival Order Model