Sadedegel

Lots to do before July 31st But we need to give a pause and publish a maintenance release for `sadedegel`. Another major release is on the way.
* **ADD**: More References
* **FIX**: Lazy loading of `AutoTokenizer` to improve `sadedegel.load()` performance.
* **ADD**: We add `validate` subcommand into `sadedegel.dataset` commandline to ensure `sents` dataset is in coherence with `raw` dataset (yes there is such a command-line which we need to document)
* **FIX**: Huge effort by dafajon and askarbozcan to fix numerous errors on our `raw` and `sents` datasets.
* **FIX**: Other minor documentation and LICENSE fixes thanks to mccakir

* **ADD**: MIT License
* **ADD**: _Reference_ section into README.md
* **ADD**: `RandomK` summarizer. `RandomK` summarizer is another baseline summarizer selecting random K (without replacement) sentences out of all sentences in document
* **ADD**: `[metadata]` section into `setup.cfg`. Extensively improved metadata content for PyPI build.
* **REMOVE**: `sadedegel.annotator`
* **FIX**: `flake8` parameters
* **FIX**: Documentation

We are finally there as a working library that can be installed and used. Obviously we are still far from being production ready.
With this first release:

* We bundled sadedeGel wheel and let others download and start using it
* pylint, flake8 and bandit integration is completed.
* We shaped the library structure (which will be changed in upcoming releases)