Towards Sharing Task Environments to Support Reproducible Evaluations of Interactive Recommender Systems