Refactor/redesign ensemble modeling

[jrcalgo]

After REINFORCE implementation, redesign the learning process and NN architecture design to leverage conceptual processes like bagging, stacking, boosting,etc. in an RL setting.

REINFORCE algorithm will provide a strong baseline in lunarlander environment. From there, we can experiment with the ensembled experts to learn using a novel policy gradient algorithm.

Extensive testing and alterations will occur in this issue's associated branch.