More effective job cancel check points

[rompetroll]

Datahub users can stop running jobs using mim jobs operate -o kill <jobtitle>. When they do, a context is canceled, and there are a couple of places in the pipeline where datahub checks the context-cancel-state and aborts (kills) job processing.

In practise, these context check points are not frequent enough. Specifically jobs using MultiSource can spend alot of processing time processing source dependencies without checking for canceling.

Datahub should do cancel checks more frequently, so that users always can kill jobs in a timely manner