openQA upstream has changed to use 'softfailed' as an actual
result state which appears in the job dict returned by the API
(previously, soft failed jobs were 'passed' so far as the API
was concerned, and the web UI inferred the 'soft failed' state
from the test module results). So we should handle this. In
Fedora usage, so far, 'softfailed' tests are basically treated
like passes. Both the overall test result and each test module
result can be 'softfailed', so handle both these.
As a future improvement we might want to consider reporting
'warn' rather than 'pass' for these, and maybe even linking
workarounds to bug numbers so we can include a bug number in
the 'warn' result template.