Explain output hierarchy when dataspaces are used

[lpellegr]

Let's assume there is a Scheduler running on a machine and a ProActive node deployed on another. A user bobot submits a task that eventually creates a file test.txt on the scratch space of the compute node. If the task was configured to transfer test.txt from the scratch space to:

• global space, then test.txt is put in defaultglobal/test.txt
• user space, then test.txt is put in defaultuser/bobot/test.txt
• output space, then test.txt is put in defaultoutput/bobot/test.txt

This difference of semantic for the output hierarchy between the first and the last two spaces seems to be desired but seems not documented.