Support for larger clusters (up to 100)

[tomponline]

Related to https://warthogs.atlassian.net/browse/PF-4669

Identify what is the current practical maximum size of a LXD cluster and then identify what are the bottlenecks that are causing this restriction to allow us to work towards supporting larger clusters.

Currently we recommend to not go larger than the size of a rack (40-50 nodes) but also have seen that larger clusters are possible (the largest in customer hands I’ve seen is 70-80 nodes).