Python 2.x and 3.x iterate dicts in different orders, which has

implications for assembling query strings. We have several tests which

spy on urlopen and return fake payloads, which were failing when run

on Python 3.
This change adds sorting to the querystring parameters before we

assemble them into a URL, and fixes up the test URLs to use those.

Ran unit tests on Python 2.7 and 3.9.