Network architecture growth schedule

[shindavid]

Appendix C of the KataGo paper describes how the neural network size was gradually increased over the course of training. We should implement the capability to do the same, and do some scientific experimentation across multiple games to determine how to best configure this automatically, with the aim of minimizing the time required to reach any given desired level of strength.