Inverse Factorized Soft Q-Learning for Cooperative Multi-agent Imitation Learning