Implement Frozen Lake environment

[doctorcorral]

This environment is useful for understanding how to solve finite MDP with perfect information using dynamic programming.
Solving this environment requires an iterative policy optimization followed directly from Bellman´s expectation equations.
The environment itself is pretty simple, it just keeps track of the position of an agent in a map, and returns respective rewards.

The implementation must be fully compatible with Frozen Lake environment implementation on Gym