Adapted Wasserstein Distances And Applications To Distributionally Robust Optimization