Make test workflow more stringent

[dfornika]

I've noticed that it's possible to get a passing GitHub Actions CI test result, even if the pipeline analysis fails. This is partly because we're ignoring failures of FluViewer within the pipeline, so a failure within FluViewer isn't enough to crash the pipeline.

But even if we keep that behaviour, we should be detecting that outputs are missing or finding other issues that should cause a test failure.