Autocat

In this release we update the ML interfacing and introduce the `CandidateSelector` class.

What's New

- `CandidateSelector` class added for choosing systems to evaluate (34)

Improvements

- Whenever an ML class depends on another, it now must be provided the instantiated object as an argument rather than the class instantiating it internally. For example, `Predictor` now takes a `Featurizer` directly rather than the parameters needed to instantiate the `Featurizer` internally. This is to provide the user with more direct control over these objects. (34)
- The `NH` and `OH` placement defaults are now 45° rather than vertical (to be more general) (35)
- Minor improvements to documentation and testing (34)

This is a hotfix that adds the required `learning` related dependencies to the `setup.py`

Initial release of the learning modules and a full-fledged documentation (including deployed site at https://aced-differentiate.github.io/auto_cat).

Modules for automating structure generation for DFT catalysis studies