Remove Python 3.6 support in Tekore version 4.0

[felix-hilden]

Some changes that would need a major version bump are beginning to accumulate. We should figure out a good set of them for the next major release.

In addition to the issues listed in the 4.0 milestone, this is what's on my mind:

Dependency management

Development dependencies can be updated at will, but we could take advantage of the major bump to update runtime dependencies. I noticed that basically no one uses Tekore on Python 3.6, which is the source of two standard library backports: dataclasses and contextvars. Essentially everyone is on 3.8, with 3.6 receiving less than 5 downloads per day for the available Pypistats history.

To clarify, nothing is forcing us to do so, but this would be a good opportunity to clean things up for future development. After improving packaging and tooling, I see no reason to start thinking about another major for a while.

What else?

[felix-hilden]

Here's the EOL plan, which sets Python 3.6 to end December this year. This seems to be both safe, since nobody will actually download this on 3.6 if we don't support it, and appropriate, since the next major is nowhere in sight and the user base for 3.6 is so small.

[felix-hilden]

The download amount has actually increased, with sometimes up to 10 downloads! But I think we'll make this change anyway, since no major improvements are introduced and Tekore 3.7 will be available for Python 3.6.