Specify ncpu/workers/processes for each parallelisation step

[ddobie]

Currently the pipeline uses the one less than the number of CPUs available for all parallel operations.

This is silly because

1. the user should be able to specify how many are used in general (e.g. when running on a shared machine)
2. it is not necessarily efficient to base the number of workers on the number of cores available.

For example, the forced fitting parallelisation is limited foremost by IO (on nimbus using >~ 10 workers slows things down), but also by memory (each worker loads an image, background map and rms map or ~2.1GB of data in total, into RAM)

We should therefore add arguments in the config file for the user to specify how many workers should be used for each parallel computation.

[ddobie]

Closed by #777