Using Inverse Optimization to Learn Cost Functions in Generalized Nash Games