Situated GAIL: Multitask imitation using task-conditioned adversarial inverse reinforcement learning