v0.1 knowledge distillation

Pre-trained Alpha Zero model on positions from KataGo selfplay. Ownership of local positions trained against KataGo predictions. Segmentation into local position computed from KataGo predictions. First local move taken from the actual game.