Neurosynchro

- More documentation updates. We now provide a sample training set and sample
pre-trained networks on Zenodo, so it's much easier to work through the docs
in a tutorial style if you'd like.
- Default `rho_Q_sign` to use the `generic` trainer in `neurosynchro
init-nndir`. This scheme works a bit better with the sample training set.
- Fix the `binary` trainer!

- Write a bunch of documentation
- Add a tool for doing test integrations to compare how well neurosynchro
fares compared to fully-detailed calculations.

- First packaged version; several micro releases to iterate on the ReadTheDocs
infrastructure etc.