Fix a performance problem

Now it only costs 9 sec. to execute UCT algorithm including 3w monte carlo game, in a single thread.

Big Refactoring and fix all known bugs

A big refactoring is done. The architecture and codes are more readable and flexible. It's convenient to experiment a variety of algorithms.
All known bugs is fixed.

It performs well on 9x9 board

v0.1.0-alpha

import gtest to do unit tests, and replace the original ugly test cod…