Function Approximation for Solving Stackelberg Equilibrium in Large Perfect Information Games