Tackling Mopion Solitaire by ranked reward reinforcement learning

This is the first/ongoing attempt to tackle Morpion Solitaire using ranked reward reinforcement learning.

For Morpion Solitaire, please see http://www.morpionsolitaire.com/

For ranked reward algorithm, please see https://arxiv.org/abs/1807.01672

The sketelon code structure is based on https://github.com/suragnair/alpha-zero-general

This is a primary trial on applying deep reinforcement learning to play Morpion Solitaire. Many places can be improved, so any suggestion is welcomed and appreciated.