A Divergence Minimization Perspective on Imitation Learning Methods