Investigate when nodes die but not enough for DR

[DomAyre]

I did some investigation where I manually restart a CCF node container:

• The node attempts to join as if it never existed
• The CCF network denies it because it believes it already has a node with the same URL

So this doesn't just work, I spoke to Gaurav from the az cleanroom team, and he said the extension is capable of detecting dead nodes but doesn't automatically re-provision them. Orchestration is deferred to a higher level process.

I see a few possible solutions to this:

1. We build a simple "orchestrator" which could just be a cron job type process which inspects the network health, and calls the scaling function if any nodes die such that the desired number is always maintained
2. We rework the az-cleanroom containers such that they do some work to re-identify as the dead node
3. We rework the az-cleanroom container such they join as a new node

[DomAyre]

Amaury confirms that we will want a separate process which monitors network health and re-provisions nodes as required.

[DomAyre]

We need a story for code upgrade on the orchestrator