Explain how leftover files and directories are meant to be interpreted if jobs fail

[LukasNickel]

If something fails, the user is left with a lot of files and directories. It took me a while to figure out what to do with that, maybe we should add to the README how to interpret these directories (e.g. running_analysis)

[vuillaut]

There is a script to cancel all jobs of a prod.
Shall we add a "clean tree" to that script and document it for the end user?
@garciagenrique