Smact

- Fixed a bug in data_loader.py so now ICSD oxidation states are available from the Element class.
- Updated README in examples folder to include description about structure prediction examples

What's Changed
* Updated oxidation states module and structure prediction module to be compatible with pymatgen v2022.0 and higher
* Binary and ternary ionic substitution functions added to structure prediction module
* Extended radii dataset added
* Updated requirements
* Example notebooks for structure prediction module included

**Full Changelog**: https://github.com/WMD-group/SMACT/compare/v2.0.2...v2.3

We have moved to the widely used and unrestrictive MIT license.

This release also includes the [new structure prediction module](https://smact.readthedocs.io/en/latest/smact.structure_prediction.html). This is a minimalist and lightweight framework for predicting new compounds based on species similarity. The implementation is inspired by [this work](https://pubs.acs.org/doi/10.1021/ic102031h) and we are looking to include other species similarity metrics in the future.

This release follows peer review of the package in the Journal of Open Source Software (see the full open review process [here](https://github.com/openjournals/joss-reviews/issues/1361)).

Most important changes include:

- Updating the README to make it more user-friendly, especially for those new to the field.
- Consolidating and improving the [examples](https://github.com/WMD-group/SMACT/tree/master/examples/).
- Updating [the docs](https://smact.readthedocs.io/en/latest/) and fixing typos.
- Renaming the `smact_test` function to `smact_filter`.

A bug in the oxidation states module has been fixed for this release. The `pair_probability` and `compound_probability` functions now work as expected.

For this release we have cut out some redundant functions, improved the docs and examples, streamlined the installation process and improved test coverage.

Installing with pip will now install all requirements, including ASE and Pymatgen. These packages are only required for some modules so alternative installation instructions are provided.

Key additions to the code in this release:
- The oxidation_states module, which can be used to assess the likelihood of metals adopting certain oxidation states in the presence of given anions (See [related paper](https://www.doi.org/10.1039/C8FD00032H)).
- `smact_test`, which applies the standard tests of charge neutrality and electronegativity order in one handy function.
- A function `ml_rep_generator`, to generate a vector representing a composition, useful for machine learning.