Change I03c8e8225e51fd80580772752c0b292987e34218 added
another notification to the list checked in
test_volume_swap_server_with_error and then change
Id587967ea4f9980c292492e2f659bf55fb037b28 relied on
checking the last notification in the list (compute.exception)
but per the note in the test, that last exception can
race to show up by the time we check the size of the
notification list which can result in an IndexError.
This change fixes the race by just waiting for the
compute.exception notification rather than the racy
_wait_until_swap_volume_error which depended on the
cinder fixture swap_error variable (which is removed
in this change).
Change-Id: I8165fcd98e11b9155640559a36ae90a38f63c0dd
Closes-Bug: #1814177