Allows dgenies-scheduler to keep going even if cluster is not reachable.

[pbordron]

In case of cluster maintenance dgenies scheduler will fails to submit jobs to cluster, but it can keep running small jobs (depending to the dgenies configuration). When drmaa submission fails because of unreachable cluster, we must put the job again the job in waiting state instead of error.

In this context, uploaded files must be stored on a storage infrastructure that is independent to cluster storage.