Alert for failing nodes to join a cluster

[njuettner]

Follow up on #3828

In order to detect this behaviour, it would be good to track errors in bootstrap controller.

So if bootstrap token cannot be refreshed we would need an alert which counts logging entries, using a counter would be the easiest thing to do.

sum by (name) (count_over_time({app="cluster-api", pod=~"capi-kubeadm-bootstrap-controller-.*"} |= `failed to refresh bootstrap token`| logfmt[5m]))

Currently on hold because Atlas is dealing with the implementation

[fiunchinho]

Hey @njuettner . Isn't that issue already covered by this alert that we added back in November?