We've seen a trend of tempest-full jobs timing out because of
tests taking a long time to complete which are not marked slow:
http://status.openstack.org/elastic-recheck/#1783405
Based on some analysis of the slowest tests:
https://ethercalc.openstack.org/dorupfz6s9qt
This change marks several tests as slow and gives reasons why
they are marked as slow - not only because they are slow, but
also because they are less common scenarios. These will still
be run in the tempest-slow job so we aren't losing test coverage
with this change.
If this makes the tempest-slow job itself take significantly
longer, we should consider running the tempest.api* slow tests
concurrently (like they are in tempest-full) and continue to
run the tempest.scenario* tests in serial. The tempest-slow job
currently runs all slow tests in serial, so this will make the
job take longer.
Partial-Bug: #1783405
Change-Id: I2a0e154ba38c7407b41b7e986a0bf9b451c65cae
b572053303