Create a new cluster before deleting the old one on the job update

[JuliaBogdan]

When updating the streaming job, we need to minimize the disruption time that comes from the Flink cluster re-creation.

One option to address it might look like blue-green deployment:

1. Flink operator receives an update request
2. A new cluster is created
3. When the new cluster is ready, the operator triggers savepoint (if JobSpec.takeSavepointOnUpdate: true) and stops the old job
4. When savepoint is taken, the new job starts in the new cluster.

An important requirement is that two identical jobs should not run simultaneously.