initiator-dictionary update causes iSCSI restarts on nodes already in map

[Paul Goins]

This is a major usability issue re: cloud expansions utilizing this charm.

As currently written, config changes on a properly configured storage-connector charm set up for iSCSI are basically guaranteed to cause iSCSI restarts.

The problem is: when clouds are expanded, they may need additional entries to initiator-dictionary. If this triggers iSCSI restarts on existing nodes, this may cause volumes in use to be unmounted, which is extremely disruptive. About the only way to avoid this is by stopping the storage-connector units across the cloud, making the config update, and not restarting the storage-connector units until each host is cleared of instances with attached volumes.

We need to be able to update initiator-dictionary while not triggering restarts for any hosts for which dictionary entries have not been updated.