ssh Add troubleshooting guide

[petersutter]

What would you like to be added:
For workers that are unable to join the cluster, there are some things that do not work as usual, e.g.

• the nodes are not listed and the host must be taken from the machine objects instead (ssh hostname from machine object as fallback #323)
• the gardener user is not there and the admin user has to be used (ssh add option to specify the worker username for the ssh command  #322) should be obsolete with [node-agent] Create gardener Linux user when nodes are bootstrapping gardener#9077
• MCM cleans up machines after 20 minutes that do not join the cluster, which makes troubleshooting hard. In this case MCM could be scaled down e.g. with

while true; do kubeclt scale deployment machine-controller-manager --replicas=0; sleep 5; done

Why is this needed: