Quantifying the Empirical Wasserstein Distance to a Set of Measures: Beating the Curse of Dimensionality