Several fixes around cachepath and commpath not cleaning up

[mplegendre]

We were no longer getting consistent cleanup of the cachepath/commpath root directory and its top level files. Several fixes related to this:

• We were doing both commpath detection and spindle_mkdir() for the commpath on the FE process. The FE isn't guarenteed to be running on a node that has a commpath. I deleted the detection on the FE.
• We had a race condition where both clients and servers were creating commpath, where it used to be just servers. If the client created the commpath it wouldn't get tracked in the server's to-be-cleaned-up data structures. Changed things so the server tracks this, even if it didn't win the race to create the commpath.
• Fixed an issue where cachepaths and commpaths were being spindle_mkdir'd with their non-realized paths. That created a disconnect from other files that were created in those directories using the realized paths. Fixed so that we always track with a realized path.