Stochastic Online Learning with Feedback Graphs: Finite-Time and Asymptotic Optimality