Pycantonese

Added
* Added the `characters2jyutping()` function for converting
Cantonese characters to Jyutping romanization.
* Added the `segment()` function for word segmentation.

Added
* Added support for Python 3.7 and 3.8.

Removed
* Dropped support for Python 3.4 and 3.5 (supporting 3.6, 3.7, and 3.8 now).

Added
* 104 stop words.

Added
* Exposed the `exclude` parameter in various reader methods
for excluding specific participants. This parameter was implemented at
pylangacq v0.10.0.

Fixed
* Allowed "n" to be a syllabic nasal.
* Fixed corpus reader not picking up the characters.

* PyCantonese now requires Python 3.4 or above.
* Adopted the CHAT corpus format, piggybacking on [PyLangAcq](http://pylangacq.org/)
* Converted HKCanCor into the CHAT format
* Switched to transparent function names
(cf. issue [10](https://github.com/pycantonese/pycantonese/issues/10)): `parse_jyutping()`, `jyutping2yale()`, `jyutping2tipa()`
* Bug fixes: issues
[6](https://github.com/pycantonese/pycantonese/issues/6),
[7](https://github.com/pycantonese/pycantonese/issues/7),
[8](https://github.com/pycantonese/pycantonese/issues/8)
[9](https://github.com/pycantonese/pycantonese/issues/9)

* Fixed the Jyutping-Yale conversion issue with "yu"
* Added ``number_of_words()`` and ``number_of_characters()`` for corpus access
* Forced all part-of-speech tags
(both in searches and internal to corpus objects)
in caps, in line with the NLTK convention