Approximating Optimal Morphing Attacks using Template Inversion