Tag: A Research Framework For Fast Prototyping Of Reinforcement Learning Algorithms