Analysis Errors Should Log to Stdout

[ax3l]

In local tests, I quite often forget to activate the right virtual environment or have just forgotten to install a few Python modules in a fresh virtual env. As a consequence, the Python scripts that we use for testing fail, e.g., with an import error.

This error fails the test that ran it, but no info besides " FAILED" is printed to stdout. After some analysis, I then find that {output_dir}/{test_name}.analysis.out contains some error message.

Long story short: we should log a line when the analysis step returns a non-zero error code. This would then be same way as we also report compile errors or crashes and speeds up error diagnosis for mortal humans like me.

[ax3l]

Proposed in #114