More consistent handling of exceptions in backend

[arielge]

There are currently some inconsistencies with how exceptions are logged and handled in the backend.

We should aim that:
A. Every exception that is explicitly raised within the backend modules should be caught upstream (usually by the app or orchestrator). For instance we do not currently catch exceptions raised by orchestrator_api.delete_workspace()
B. Each such raised exception should have an informative message, and this message should be logged where the exception is caught.