Resumable-Checkpointable tasks

[bpiwowar]

Some tasks (e.g. using checkpointing) can be restarted if stopped by a time limit (e.g. with SLURM). It would be good to be able to restart them, so that this can be properly handled:

• A restarting tasks must derive from ResumableTask (that derives from Task)
• Non resumable tasks directory can be cleaned up before starting, but not for resumable ones – this includes dynamic outputs in particular
• For schedulers that support this, a new exit error (Timeout) has to be created and handled

[bpiwowar]

Issue #122 should be fixed since the logic will be much easier to implement: when a job fails, if this was because of a timeout, then restart