Stop testing Python 3.3 in Travis

[jmbowman]

Summary of changes

The Python 3.3 job in Travis failed in the last build on master because the new wheel release won't install on it. We planned on dropping 3.3 support soon anyway, so just stop testing it on Travis to avoid further job failures. We'll release one more version that officially supports Python 3.3 while incorporating all the latest fixes from master, and then the next one will fully remove support for it.

Pull Request Checklist

• News fragment added in changelog.d. See documentation for details

[pganssle]

This is already done in #1342 and is deferred to the next major release.

If installing wheel is breaking Travis, that is a bug in wheel, which should be using Requires-Python.

[pganssle]

Actually, my bad, it looks like pip doesn't have a version that supports Python 3.3 and Requires-Python.

I think it's time to land #1342.

[jmbowman]

I talked to @jaraco about this at the PyCon sprint earlier this morning, and our tentative plan was:

• Merge this PR so we have clean Travis builds again.
• Release 39.2.0 today with all the fixes already in master or ready to go in today (things that already passed Python 3.3 tests before the new wheel release came out).
• Merge Drop support for EOL Python 3.3 #1342 immediately after that, so the next release would be 40.0.0 without Python 3.3 support.

Would you like to handle this differently? We could immediately merge #1342 and go straight to 40.0.0, just thought it would be nice to make the latest fixes available to any remaining Python 3.3 users before doing that.

[pganssle]

Hm, this is a tiny bit complicated. I was hoping that we could just pin wheel at 0.30.0 until the 40.0.0 release (and just revert the single commit that does that), but it seems there are some more complicated dependencies in the test suite.

[pganssle]

I think I have a more surgical PR in #1373.

I made it a single commit so it can be reverted as part of #1342.

[pganssle]

Eh, #1373 seems like it's not quite working and I don't think it's worth being fiddly about. This approach works fine.