[Node Role]: improve handling of failure during the bootstrap

[vladzcloudius]

HEAD: cee4c41

Description

When the Role waits for a node to bootstrap (https://github.com/scylladb/scylla-ansible-roles/blob/master/ansible-scylla-node/tasks/start_one_node.yml#L10) it would wait until timeout (which may be very long).
If the bootstrapping failed while it waits the Role would still wait till the timeout wasting precious time.

We should improve this and make the Role catch the situation when bootstrapping failed and break out immediately.

[vladzcloudius]

One way to do it is to check the state of the scylla-server service on every iteration.
If bootstrapping failed - it will not be "running".