Neighbormodels

This update accounts for double counting correctly by dividing all interaction model coefficients by 2.

This update fixes the incorrect counting of neighbors with `i=j`.

New feature

`distance_filter` is a new argument for the `build_model` function that allows the user to pass a dictionary containing pair distances. Any pairs whose separation distance is given (within the tolerance of the distance bins) will be kept, all other pairs will be removed. If `distance_filter` is not specified, then the standard behavior of including all pairs within the distance `r` is assumed.

Refactoring

Grouping and ranking the distance bins now happens as part of the `count_neighbors` routine. This information is then joined to the main neighbor counts data frame when building the model coefficients data frame with the `build_model` function.

Some functions and attributes were renamed to be more consistent with refactorings that I implemented in [sPyns](https://github.com/datamaterials/spyns).

All code has been passed through the Black code formatter.

The following is a proof-of-concept prototype of the `neighbormodels` application. Using the primary functions within each module, a user is able to build a lattice model with pairwise neighbor interactions.

The API **will** change significantly from the current form. While you can already compute some useful things with this code, if you use this release for anything meaningful, pin the version tag in your project, otherwise an update will most likely break your code.