Author-rank

Fixed
- An issue that caused AuthorRank to fail when documents
without authors are accidentally provided as part of the
set of documents. Additional checks have been added to
`author_rank/graph.py`.

Fixed
- An identified issue that resulted in disconnected (i.e. non-co-authoring)
authors to be connected in the AuthorRank graph. Several code changes
were introduced to address this issue.

Changed
- The progress bar functionality such that it indicates progress on the
`.fit()` function across both the graph creation and scoring of authors
(previously the progress bar only covered the former).

Fixed
- A typo in `author_rank/__init__.py` that referred to the library as
`NetworkX` and not `AuthorRank`.

Changed
- The manner in which users interact with the library to more
closely mirror the conventions of the [scikit-learn](https://scikit-learn.org/)
and [NetworkX](https://networkx.github.io/) libraries, as described in
[this issue](https://github.com/adidier17/AuthorRank/issues/10).
- Example code and Jupyter notebooks to align with the changes
described in the above line item.

Added
- A warning to users when attempting to fit AuthorRank with a
document set that has a single author. AuthorRank requires at
least 2 authors in the document set.
- A warning in the case that users attempt to call `top_authors`
prior to `fit`. The AuthorRank approach must first be fit
to a set of documents before returning the top authors.
- A test to ensure that data processing occurs within a specified
time bound (in seconds).

Fixed
- A bug whereby the incorrect list of authors per document
was being processed into the AuthorRank graph.

Added

- A progress bar as an optional argument for creating the `graph.create`
and `score.top_authors` functions, which provides users an indication of
how far along the processing is on the input documents.

Fixed
- An ZeroDivisionError that would result in the top_authors calculation
failing in some cases. A dataset that results in this case has been
added to the testing suite and a fix developed.

Changed
- Libraries required listed in `setup.py`.

Changed

- Updates the normalization of scores in `top_authors` to a pure Python
approach, removing the `numpy` and `scikit-learn` requirements.