cluster-manager should be able to perform operations on multiple nodes

[mapuri]

At present cluster manager operations like discover, provisions or decommission one node at a time. This was done to keep failure handling simple and get started with trying cluster workflows

But there is obvious need for cluster-manger to be able to perform operations on multiple nodes together for ease of management.

Design considerations:

• how does this interact with node role/host-group management (#69) ?
• how to handle provision failures?
  • is it ok to cleanup all nodes ( #85) i.e. have a all or none semantic. This seems more deterministic from user perspective.

[mapuri]

closed by #106