Fix failed domain non-retrying.

If a domain resolution ends in STATE_BAD (e.g. a SERVFAIL), the code
never retries that domain's resolution again, though it clearly means
to. It marks the domain as bad and says "holding off", but when the
bad domain marker times out, it simply gets again the old resolution
result and repeats marking the domain as bad.

This change removes the domain resolution result when the domain is
marked as bad, so that once the bad domain marker times out, there is
no existing result and a new domain resolution is started.

bin/petemaild

@@ -1194,6 +1194,7 @@ Log::print('debug', "  -> domain $domain not resolved, adding to resolver queue"
         if ($r->failed()) {
 Log::print('debug', "  -> resolution of $domain failed, holding off");
             push(@{$self->{baddomains}->{$domain}}, $now);
+            delete($self->r()->results()->{$domain});
             # have another go at the message later
             push(@messages_to_requeue, [$msg, $now + MIN_DOMAIN_BAD_INTERVAL]);
             next;