Rebalance request may ignore recovery nodes if config has not synchronized

Delta recovery nodes are synchronized as part of the cluster configuration, which by default is an eventually consistent operation.

This means that it is possible for one node to be informed of a node recovery (via REST or other means) and for the orchestrator node to subsequently rebalance before the node recovery has been acknowledged on the orchestrator. This results in the node that the user has expected to be recovered (since the REST call responded with a 200 status code) to be removed as part of the rebalance.

Either the rebalance API needs to be adjusted to allow the user to specify which nodes specifically are to be recovered, or for the janitor to force a configuration synchronization before deciding which nodes to remove as part of the rebalance.