Since we aim at accelerating some of our functions, one way to do this is to use parallel computing. We already have one function (restoration.cycle_spin) which is using dask.delayed and several threads, there is also the attempt to parallelize segmentation.slic in #3120 which I plan to revive, and probably other functions will have in the future the possibility to use one or more jobs/workers (threads or processes). Therefore, I'm opening this issue so that we can decide what is the best terminology for this parameter. Should it be

• num_workers (as in cycle_spin, and as in dask)
• n_jobs (as in scikit-learn and in joblib)
• something else? (probably not, it's unfortunate enough that joblib and dask have a different convention)
  I think both are fine, the choice should be more about consistency with the rest of the ecosystem: do we want to be more consistent with our backend, or with big brother scikit?