This adds the ability for the ceph-osd osd-directory.sh script to
handle existing deployments that place data in hosts via CRUSH and
modify those deployments to place data in racks instead. The
existing data remains intact but is redistributed across the new
rack-level failure domains by updating the CRUSH map and assigning
new rules to existing pools.
Change-Id: Ida79f876d0cae3d99e796e4de1aac55a7978986c
810026f342