While Gram4 tries to cancel a job at the local resource manager, if the job
is currently running in the local resource manager when a client calls
destroy() for the job, it never interrupted a currently running staging if
a job was in the middle of staging when the client calls destroy() for the
job.
In fact it doesn't even wait for the staging to be completed before starting
the clean up steps.

Solution:

In 4.2 Gram interrupts stageIn and stageOut by destroying the transfer
resources before the cleanup steps are started. fileCleanUp is not
interrupted, because this would have to be performed anyway in termination.