Samewords

Changed
- Improved Unicode support. The script now fully supports Unicode 10 and
correctly recognizes word characters in any language.

Added
- Separate documentation on
[https://samewords.readthedocs.io/en/latest/](https://samewords.readthedocs.io/en/latest/).
- ``--update`` and ``--clean`` features (along with a ``--annotate`` argument,
but that is the default mode).

Added
- Full unicode support.
- Configuration option to modify the set of punctuation characters.

Solved
- `thinspace` and `\,` is now tokenized correctly and does not break the
annotation [26].
- Bug with unwanted nested sameword annotations when more than two `\edtext{}{}`
macros contained the same phrase of more than one word [21].
- `context_distance` is now also actually included in the update of the settings
during parsing of user configuration file [22].
- Words with intervening comments are annotated correctly [21].
- Consecutive sameword context matches now actually all get annotated [20].

Changed
- Make the annotation more robust to `\edtext{}{}`s and `\lemma{}`s where the
first or last (tokenized) word is not a word (e.g. `\edtext{ %\n
words}{\lemma{ %\n word}\Afootnote{note}}`).
- Reverse the change from 0.2.6. I have realized that I don't want to include
test assets in the binary distribution that you get from pypi because it's not
the idea that you run tests on that. If you download the source distribution,
either from pypi or Github of course you get the test assets there too.
- Determination of Macro.name includes all types of whitespace [19].

Changed
- Updated `setyp.py` with `data_files` to actually include what has already been
included in the MANIFEST in bdist build too(!)

Changed
- Updated the MANIFEST to actually include the assets.

Added
- The amount of words compared in the context is now configurable.
- MANIFEST.in file for including test assets.

Changed
- The default context comparison distance is reduced from 30 to 20.