"unmanage" command

[bertschinger]

There should be a command like halo unmanage <resource> or halo unmanage <host> that causes halo to continue to observe a resource's state, but never take any action to manage that state.

Such a resource should be identified in the halo status output so that it is obvious to the sysadmin that the resource is not being managed.

In addition, halo manage ... should be implemented to reverse the unmanage operation.

The state of being unmanaged should be stored persistently (using the mechanism implemented in #9) so that an unmanaged resource stays unmanaged after the halo process restarts.

[bertschinger]

Design ideas:

• struct Resource should have a boolean managed field
• since it doesn't make sense to manage a child resource if any ancestor of that resource is unmanaged, setting the managed flag on a resource should recursively update the flag on all of its children
• If a resource group is partially unmanaged, then the state of any unmanaged resources in that RG should have no bearing on whether the RG as a whole is considered to need failover. For example, an unmanaged child resource being down or timing out shouldn't cause the RG to failover if the (managed) parent resource is OK
• how are resources actually identified? Currently they don't have an easy to use "ID". An ID field may need to be added to the struct and to the config file.