A Bayesian optimization approach to find Nash equilibria