use fsGroup in MGM, reduce init container
For now this is an incremental improvement that removes one initContainer for the MGM. It combines all mkdir operations in one container that runs with runAsUser: 2 (thus creating the required subdirs with the right user ownership), and sets fsGroup: 2 for the MGM pod (thus ensuring that the group of files in mounted volumes is correct). This also removes the need for any chowning, which also speeds up MGM startup.
I compared the ownership of all files in the /var/eos and /var/log/eos volumes before and after. The only difference is that some files or directories which were (accidentally?) created with root:daemon ownership are now daemon:daemon. This fixes a small possible risk that the unprivileged daemon might not be able to write to a location where it should.
I think this is as much as can be (easily) done for #83 for now, and it will either help or be required for !47 (closed)