Malaffinity

-------------------

* Remove scipy (and numpy) as a dependency. Pearson's correlation code is now in
``malaffinity.calcs`` and stdev checking is handled by the ``statistics`` module.
* Use ``lxml`` for XML parsing, instead of the default ``html.parser``.
* Add return types for components inside the return tuple into the docstring.

-------------------

* Change 'base user has been set' testing to also check if ``self._base_scores``
has been set as well.
* Use ``zip`` to create the ``scores1`` and ``scores2`` arrays
that calculations are done with.
* Check if the standard deviation of ``scores1`` or ``scores2`` is zero,
and thrown an error if so.
* Use ``scipy.asscalar`` as opposed to ``.item()`` for numpy.float64 => float conversion.

-------------------

* Better handling for numpy.float64 => float conversion.
* Update docstrings to include types.

-------------------

* Don't count rated anime on a user's PTW. MAL didn't count this,
so our affinity values were a bit off when a user did this.

-------------------
* Konnichiwa, sekai!