Imitation Learning of Deformable Object Manipulation with Entropy-maximizing Dynamic Policy Programming