dns: adjust test @dns_dnsmasq_kill_ratelimit for change in 1.22
When dnsmasq process keeps crashing, NetworkManager will restart it, but also ratelimit the starts.
Before 1.21.1, when ratelimiting hits, NetworkManager would write the real name servers to /etc/resolv.conf, instead of 127.0.0.1. The aim ws to workaround the issue. However, in this case something is really wrong already. When dnsmasq keeps crashing, there is no solution except letting the user investigate the issue. Just pretending that for a moment all is good, until we retry starting dnsmasq again, is not helpful.
This will change in 1.21.1 [1]. We of course still ratelimit the restarts, but we don't bother "fixing" the situation. Yes, name resolution won't work, but that seems preferably, as it lets the user become aware of the problem.