Replace all uses of six.move by the Python 3-only equivalent

[foolip]

To improve confidence that this will work in Python 3.6, all of the
unique import statements were tested in Python 3.5.5 and 3.6.8:

from http.client import HTTPConnection
from urllib.parse import parse_qs
from urllib.parse import parse_qs, quote
from urllib.parse import parse_qsl, SplitResult, urlencode, urlsplit, urlunsplit
from urllib.parse import unquote
from urllib.parse import urlencode
from urllib.parse import urlencode, urlparse
from urllib.parse import urlparse
from urllib.parse import urlunsplit
import urllib

[foolip]

I've looked through the before/after results and don't think I've caused any of the changes in webdriver for Firefox and Safari:
https://wpt.fyi/results/webdriver/tests?diff&filter=ADC&q=is%3Adifferent&run_id=5676157900947456&run_id=5656274618286080
https://wpt.fyi/results/webdriver/tests?diff&filter=ADC&q=is%3Adifferent&run_id=5676288159252480&run_id=5181381359435776

But there are "setup error" failures which might be how a failure manifests, so I'm going to have to take a closer look.

[foolip]

I rebased to trigger CI again, here are the before/after of Chrome/Safari:
https://wpt.fyi/results/webdriver/tests?diff&filter=ADC&q=is%3Adifferent&run_id=5692740903895040&run_id=4823262422892544
https://wpt.fyi/results/webdriver/tests?diff&filter=ADC&q=is%3Adifferent&run_id=5733793677705216&run_id=5109353482813440

Comparing that with the diffs in #28735 (comment) I'm now confident these aren't regressions of these change. There doesn't seem to be any test that passed both times without the changes and then failed both times with the changes.