Implement optimality tightening

[Learning to Play in a Day: Faster Deep Reinforcement Learning by Optimality Tightening](https://arxiv.org/abs/1611.01606) potentially speeds up Q-learning by an order of magnitude! Apparently not too hard to implement either.

Start by reading the linked paper, “Learning to Play in a Day: Faster Deep Reinforcement Learning by Optimality Tightening,” and then locate the repository’s Q-learning implementation. Done means implementing the proposed optimality-tightening approach and demonstrating whether it produces the claimed speed improvement.

Written by the indexing model from the issue text.