Node install stalls because of large retry count

[ctrox]

On RKE2 we have observed that the machine-provision pod can sometimes be stuck for hours due to the very large retry count of 4500. This mainly seems to happen in retrieve_connection_info, which by the way does not exit 1 even once it is done with all the retries.

Regardless of the actual cause making retrieve_connection_info fail all the time, wouldn't it make sense to have a more reasonable RETRY_COUNT here? This would cause the provisioning to fail faster and retry by creating a whole new machine.