Mirror Descent Strikes Again: Optimal Stochastic Convex Optimization under Infinite Noise Variance