Textnets

- Adds Catalan, Macedonian and Russian language models.
- Significantly speeds up backbone extraction by implementing the disparity filter integrand in Cython. (If the compiled extension
cannot be loaded for some reason, it falls back on an interpreted function.)
- [PyPI](http://pypi.org/project/textnets) *should* now receive binary wheels for Mac, Windows and Linux (via GitHub Actions) to ease installation on each platform.
- Improved type annotations.
- Update several dependencies.

* Improve the handling of edge cases when initializing the `Corpus` and `Textnet` classes, such as empty data being provided.
* Added ability to run the tutorial in [the documentation](https://textnets.readthedocs.io/) interactively using [thebe](https://thebelab.readthedocs.io).
* Update to spacy 3.1 and bump other dependencies.

* Adds `Corpus.ngrams` method as alternative to `Corpus.noun_phrases`. This is useful when working in languages that do not have noun chunks, such as Chinese. The tutorial now also mentions this as an alternative to `tokenized` and `noun_phrases`. (Closes 26.)
* Fixes a bug in `Corpus.from_files`. (This bug was probably part of the issue in 28.)
* Introduces HTML representations of core classes for nicer integration in Jupyter notebooks.
* Updates several dependencies (`pandas` and `leidenalg` among them).

* Migrate continuous integration testing from Travis to GitHub Actions.
* Continuous integration tests now run for MacOS and Windows too.
* Update to Spacy 3 and bump other dependency versions.
* Improvements to documentation.
* Handle dependencies and build project using Poetry (PEP 517 and 518).
* Remove deprecated command-line interface.

* Python 3.9 compatibility!
* Updated documentation with conda-forge installation option.
* Bump versions for numerous dependencies.

- Add `cairocffi` dependency and update installation docs.
- Bump `leidenalg` dependency to version 0.8.1.