The MineRL Competition on Sample Efficient Reinforcement Learning using Human Priors