In test cases, use gmtime to avoid timezone & daylight savings issues

[dereksz]

This is just a fix to a few test cases; it does not make any changes to the pip source code.

I'm in Sydney, Australia, and just started looking at the pip code for the first time.

First thing I do is clone and run the test via nox. I found test cases failing because of a difference in time stamps found on log messages. Looking further, I see that someone had already had the sense to use time.timezone to adjust the timet style timestamps into local times. However, this adjustment fails when in Sydney as the baked-in date used by the testcases (1547704837.040001 / 2019-01-17T06:00:37,040) falls within daylight savings here, and so the formatted time becomes "2019-01-17T07:00:37,040" - an hour ahead.

I've had enough pain with python, timestamp and time-zones to NOT want to have to make pip a test-bed for all of python's time-zone shortcoming, so the fix in this PR is to have these test cases monkey-patch the root logging.Formatter to use converter=time.gmtime, and to remove the earlier time.timezone adjustments. This now allows the tests cases to pass in southern hemisphere geographies that observe daylight savings.

[uranusjr]

Should this be in a fixture instead? Modifying the global state in test seems wrong to me.