Specify multiprocessing start method
I noticed that we weren't getting the expected CPU usage for jobs I submitted and tracked this down to a (still unknown) difference in lalsimulation behaviour when using the fork start method for multiprocessing. I've also seen badness when using the spawn method on MacOS.
To address this, I've manually set the start method to forkserver and verified that this works on an M1 MacBook Pro and on several nodes at CIT.
I know it is late in the day, but I'd like to get a version of this in the next release as this issue is really hammering review runs and will presumably also impact online/production runs.