Drop Python 2 support

[Harmon758]

With Python 2.7 reaching its EOL at the end of the year, I'd like to start planning to drop support for it.
It would increase maintainability and reduce a lot of the overhead required for compatibility, as well as allowing improvements that would be difficult to implement while supporting Python 2.

Specifically, I want to look into adding asynchronous support with asyncio (#732), i.e. async/await syntax, and maybe with aiohttp (or requests 3 if that's released soon), especially to replace the current threading method for streams.

I think the current plan should be for Tweepy v3.9 (v3.9 was released without noting that it would be the last version to support Python 2 in its changelog and release notes, so a v3.10 is planned) v3.10 to be the last version to support Python 2.7, and then for v4.0 to be Python 3.5+ (3.5 is also EOL now) 3.6+.

[hugovk]

The plan sounds good. Python 2.7 installs of Tweepy are relatively low for Tweepy.

Here's the pip installs for Tweepy from PyPI for July 2019:

category	percent	downloads
3	73.34%	83,887
2	25.93%	29,655
null	0.73%	836
Total		114,378

Source: pip install -U pypistats && pypistats python_major tweepy --last-month

category	percent	downloads
3.6	37.23%	42,582
2.7	25.89%	29,611
3.7	18.38%	21,022
3.5	15.50%	17,734
3.4	2.08%	2,381
null	0.73%	836
3.8	0.09%	98
3.3	0.06%	69
2.6	0.04%	44
3.9	0.00%	1
Total		114,378

Source: pip install -U pypistats && pypistats python_minor tweepy --last-month

[hugovk]

Please see PR #1255.

[Harmon758]

Since Python 2 has been EOL for almost a year now and PyPI Stats shows Python 2 downloads averaging at about 5% or less, and with Tweepy v3.10.0 released now, I've gone ahead and made a branch and PR #1482 to drop support for Python 2.

Note, Tweepy v3.10.0 will also be the last version to support Python 3.5, since it's also EOL now.
PyPI Stats shows downloads for it averaging at even less, about 2-4% or lower.