Move ls to run-worker, one level only

[bethac07]

Resolves #214

Currently available as bethcimini/distributed-cellprofiler:2.2.1_4.2.8_fixatt4 ; had one small-scale test but would be great to have a larger test before we pull!

[ErinWeisbart]

Thanks Beth!
Do you mind explaining briefly what the other fixes we tested were (since I know not all of them worked) so that we have a record here in case this needs to be revisited again for some reason?

[bethac07]

Sure!

Previously, inside cpworker, we os.listdired the bucket root and then subprocess lsed one level down. Exactly WHY that very strange combo is lost to time inside my brain, though possibly a more diligent search of 2016-era issues/commits might turn up if other things were tried.

Fully removing any ls of the mounted system was the only thing that caused bucket mounting to break(per my understanding - I made the containers but did not personally test them). Keeping either an os.listdir OR subprocess call to ls inside the cpworker script at only a single level worked to mount the bucket. So did a two-level ls in run_worker (aka ls /home/ubuntu/bucket/*/) . But since the cpworker script runs every job, and the run_worker script runs only on container mount, this solution (ls /home/ubuntu/bucket/ ) generates the minimal "touches" (and therefore costs) that still preserve functionality.