correct EDNS responses (#96)

Tests updated as well and all the middleware. And Prometheus renamed to
metrics (directive is still prometheus).

middleware/metrics/README.md

@@ -0,0 +1,34 @@
+# prometheus
+
+This module enables prometheus metrics for CoreDNS.
+
+The following metrics are exported:
+
+* coredns_dns_request_count_total
+* coredns_dns_request_duration_seconds
+* coredns_dns_response_size_bytes
+* coredns_dns_response_rcode_count_total
+
+Each counter has a label `zone` which is the zonename used for the request/response,
+and a label `qtype` which old the query type.
+The `response_rcode_count_total` has an extra label `rcode` which holds the rcode
+of the response.
+
+If monitoring is enabled queries that do not enter the middleware chain are exported
+under the fake domain "dropped" (without a closing dot).
+
+Restarting CoreDNS will stop the monitoring. This is a bug. Also [this upstream
+Caddy bug](https://github.com/mholt/caddy/issues/675).
+
+## Syntax
+
+~~~
+prometheus
+~~~
+
+For each zone that you want to see metrics for.
+
+It optionally takes an address where the metrics are exported, the default
+is `localhost:9135`. The metrics path is fixed to `/metrics`.
+
+## Examples