Redid the API to make it much neater and removed a lot of unnecessary features, added in the option to ensure learner's bid uniqueness from mutation. Otherwise not much else changed functionally, the code is just a lot nicer to look at now. Future releases will likely include helper functions in the utils file for stuff like running TPG in a gym environment and transforming of inputs, etc. Also an optional shared memory module will be implemented too, as in https://www.researchgate.net/publication/332327816_A_Model_of_External_Memory_for_Navigation_in_Partially_Observable_Visual_Reinforcement_Learning_Tasks .