💫 Tidy up dependencies

[ines]

Description

We're always looking to drop unnecessary dependencies. While spaCy's dependencies are generally quite healthy (and mostly consist of our own libraries), we were still able to identify several libraries that we didn't actually need – and which dragged in a range of other dependencies.

We've already removed six (6d2c85f), html5lib (mostly a workaround for a six issue) and ftfy (ce6071c, ac88c72).

The PR also replaces requests with urllib / urllib2. After inspecting the code again, it turned out that the library was only used in two places in the CLI and only to fetch JSON files. The new solution works fine on Python 3 – the Python 2 compatibility workaround still needs to be tested in detail.

We'd also like to upgrade msgpack-python to the renamed msgpack (see here for details). The way the library was renamed was kind of awkward, and a recent update caused various problems for spaCy users. So we want to be careful here. Another difficulty is that this change needs to happen in sync with Thinc, which also depends on msgpack-python. Otherwise, it will result in clashing versions again.

One option to solve this could be to only require msgpack via thinc. This way, spaCy won't have to specify a separate dependency, and we won't have to worry about both msgpacks being out of sync.

Types of change

enhancement

Checklist

• I have submitted the spaCy Contributor Agreement.
• I ran the tests, and all new and existing tests passed.
• My changes don't require a change to the documentation, or if they do, I've added all required information.