Fix excessive warnings about unknown poll status

Unknown pollstatus was being logged when we had PollStatusUnknown since
it is represented by NaN which is not comparable. This didn't effect
metrics, but was noisy.

pkg/pollers/host.go

@@ -210,13 +210,13 @@ func (s *Host) Collect(ch chan<- prometheus.Metric) {
 	s.PathReachable.Collect(ch)
 
 	//Latency
-	switch s.pingStatus {
-	case PollStatusUnknown:
-		s.PingLatency.Set(math.NaN())
-	case PollStatusFailed:
+	switch {
+	case s.pingStatus == PollStatusFailed:
 		s.PingLatency.Set(math.Inf(1))
-	case PollStatusSuccess:
+	case s.pingStatus == PollStatusSuccess:
 		s.PingLatency.Set(float64(s.pingLatency / time.Microsecond))
+	case math.IsNaN(float64(s.pingStatus)):
+		s.PingLatency.Set(math.NaN())
 	default:
 		s.log().Warn("Unknown PollStatus value returned")
 		s.PingLatency.Set(float64(s.pingLatency / time.Microsecond))