Envoy /stats/prometheus?filter works only with statsd metrics

```
curl 'localhost:9901/stats/prometheus?filter=runtime.load_error'
# TYPE envoy_runtime_load_error counter
envoy_runtime_load_error{} 0
```

I was expecting `envoy_runtime_load_error` to be used when filtering metrics on the prometheus stats endpoint, but filtering on it returns nothing. Only statsd `runtime.load_error` works.

This forces a weird mapping on the monitoring side. We have to maintain a map of statsd:prometheus metric names or re-implement the reverse transformation on the monitoring side.

I'm willing to work on this if it's considered a bug and not expected behaviour. What I think is happening is that the filtering is done before the statsd->prometheus metric transformation -- and should be moved after.