Add options for max memory, n_jobs

[mschubert]

One of the ideas of the option-based configuration of clustermq is that code should be portable between compute environments, e.g. when moving to a different HPC scheduler.

If moving from HPC to local parallelization (even on a heavy duty server), resources will be much more limiting.

For a multicore setup, it would be nice to provide options for setting limits on total concurrent jobs (defaulting to cores/available cores) and memory (defaulting to memory available on that machine)

This way HPC scripts would not need to be adjusted, but could run (at reduced capacity) on multicore environments