Batch MCTS, Part 1

It’s not straightforward to get good performance out of AlphaZero’s Monte Carlo Tree Search. Search wants to perform inference for one game state at a time — the new leaf state visited at the end of each game simulation — and the results of that infe...