Cluster manager: use update to wait for cluster to start

[dandavison]

This PR changes our cluster manager sample as follows:

• Previously, the workflow was started, but the "cluster itself did not start" until a signal is sent in.
• Now, the workflow automatically does the "start cluster" operations on workflow start. An update is available for the user to wait until the cluster is ready to be used.

The reason for this change is that it highlights two styles of update usage:

1. The update accepts some input and does something with it (assign nodes to cluster, or execute an activity in general)
2. The update is used to wait until a certain stage (cluster started, early return available, etc)

In case (2) the workflow should generally be written such that it will orchestrate its work whether or not the update is sent, and that's what this PR does.

Also:

• switch to use @workflow.init
• Use hash(node_id) % 5 instead of requiring node IDs to parse as ints