What’s new
This minor version adds one interface change: you can now write `guess(mode=maximize)` (resp., `guess(mode=minimize)`) as a prettier synonym for `guess(mode=maximize())` (resp., `guess(mode=minimize())`, that is, for optimising over the values themselves (`key=lambda x: x`). See the [updated documentation](https://github.com/aeporreca/nondeterminism#optimisation) for an example.
Besides, the code was simplified (there was actually no need for a stack of interprocess queues, when you can just exploit the call stack!) and reorganised.