They are really two workarounds, even though the bandit one is more likely
to stay.
- bandit: exclude 1.6.0, which has a regression in the handling of patterns.
A fix is in place in bandit master, so newer versions should work.
The version is excluded there because is not handled by global requirements.
- stestr: exclude the only py3 test. The discovery seems to work with
the current stestr master (most likely thanks to
https://github.com/mtreinish/stestr/pull/241), but a new version does
not seems in sight, so let's disable the test for now.
Change-Id: I041b5cd41b089b2c1dc043f1e252c1f6b3c29db3