Check Python process error handling

[johann-petrak]

This is relevant when a pipeline is run e.g. in GCP where an exception from the python plugin will not cause the termination of the whole execution. So if the Python process started by one duplicate dies, the plugin still gets run for more documents.
Currently, we do not re-start the process automatically if it died.

• in theory, the python process should not die if the plugin python code throws an exception
• are there any errors that should/could terminate the python process?
• note: syntax error should get checked even before the actual process gets started (separate process for checking that!)
• maybe we should have a parameter to enable auto-restarting the python process? Maybe make it a number: restart that many times at most

In any case, would be good if we test this and document exactly what is happening.