API does not provide facilities to communicate job and adhoc hostnames

Since #50 (closed), there's no way for scord (or scord-ctl) to associate a particular adhoc_storage instance with the node allocation it should use. Since this can change dynamically (i.e. for storage instances that survive during several jobs or due to a malleable increase/reduction in the number of compute nodes), it makes sense to add this information in the adhoc_storage_context instead.