Windows RBD service issues
This  PR fixes a few issues affecting the newly introduced Windows RBD service:
- New rbd mappings are now started by the ceph service, thus no longer being tied to the user session
- Fixed the block count overflow for image larger than 2TB
- Added a “--non-persistent” map option
- Limited the scope of the commands executed by the Ceph service
- Added configurable service start and image map timeouts. Also fixed an issue where the service would hang indefinitely when failing to remap images
- Changed the way in which the service handles remap failures. Previously, it was getting marked as stopped but already started daemons kept running. Now, by default, the service will keep running unless the “--remap-failure-fatal” option is set, in which case it will cleanup existing mappings before shutting down.
- Fixed a few error propagation issues
Those changes will have to be backported to Pacific.