Min-Max Optimization without Gradients: Convergence and Applications