Bartłomiej Mielnicki

@mesutyang97 Do you think it is okay to merge this code? I'm not sure how much those failed featurization tests are a concern (I'm not familiar with featurization at all...

`self.rnd_agent_pair` in tests is the greedy human model and it produced a different trajectory than before (because there were cooking any soups). Because of that featurization changed. This explain how...

Updated version with python visualizations: https://colab.research.google.com/drive/1xFNnun0Ykob71cP6oEi8Y9JKQUYXYChD?usp=sharing#scrollTo=s8fYj-bF2Z3R Some notes: 1. There is no need to rush with a review of this as https://github.com/HumanCompatibleAI/overcooked_ai/pull/53 should be merged first anyway. Mow the code...