Zero-play

Track a human players wins and losses, and also plot AI strength as you increase search iterations.

Performance improvements in 33, multiprocessing, and 35, tail recursion. Plus a minor fix to review mode in 32.

Main feature is restoring old `play` command as `zero_perf` to find performance problems in game code (23). The rest were display improvements, like moving controls out of the graphics scene (28) and disabling moves (22). Full list is on the [milestone].

[milestone]: https://github.com/donkirkby/zero-play/milestone/3?closed=1

First usable version, with a working review function to see which moves were the weakest in a game.

* Display game rules, for users to learn a new game.
* Adjust AI strength based on wins and losses.

See a full list of changes in the [milestone].

[milestone]: https://github.com/donkirkby/zero-play/milestone/2?closed=1

Add a graphical user interface for the included games, and a base class for other libraries to add interfaces for extra games.