“Computing certain regularized equilibria in two-player zero-sum games can be treated as perfect-information problems because they do not possess the non-correspondence problem found with Nash equilibria.”