Provide an variant of Open AI's Gym

[StoneT2000]

Dimensions can actually replicate Open AI's gym and make it much more accessible by machine learning enthusiasts in other languages.

There are two ways of approaching this:

1. In particular, we can actually just import Open AI's gym in a custom design in python that interacts with agents through I/O (while I/O is a little slower, the time spent at this I/O step is insignificant compared to the ML training and testing being done). The agents can still be in python if they want, they will just need to serialize their data appropriately according to the spaces and the custom python design will have to de-serialize appropriately.
2. We can create a dimensions framework based Design that effectively has the same feature set as the gym. The Design will become the gym, and it will allow for environments to be loaded into a design for use. The Design will interact with agents through by sending them observations, rewards, done status, and info, the same way the Open AI gym does. It will probably also inherit the idea of spaces which dictate the type of observations and actions available.