Add prometheus metrics

[mozzieongit]

This adds support for Prometheus metrics, exposed using libevent's HTTP server
implementation. It is feature gated with --enable-prometheus-metrics.

The statistics export and socket code is based on code from remote.c.

Here is an excerpt from the exported metrics:

nsd_queries_total{server="0"} 27
nsd_queries_total{server="1"} 22
nsd_queries_by_type_total{type="A"} 42
nsd_queries_by_class_total{class="IN"} 49
nsd_queries_by_opcode_total{opcode="QUERY"} 49
nsd_queries_by_rcode_total{rcode="NOERROR"} 48
nsd_queries_by_transport_total{transport="udp"} 25
nsd_queries_with_edns_total 49
nsd_queries_with_edns_failed_total 0

nsd_connections_total{transport="tcp"} 24
nsd_connections_total{transport="tls6"} 0

nsd_xfr_requests_served_total{xfrtype="AXFR"} 0
nsd_xfr_requests_served_total{xfrtype="IXFR"} 0

nsd_queries_dropped_total 0
nsd_queries_rx_failed_total 0

nsd_answers_tx_failed_total 0
nsd_answers_without_aa_total 0
nsd_answers_truncated_total 0

nsd_time_up_seconds_total 970.392208

nsd_size_db_on_disk_bytes 0
nsd_size_db_in_mem_bytes 0
nsd_size_xfrd_in_mem_bytes 1163464
nsd_size_config_on_disk_bytes 0
nsd_size_config_in_mem_bytes 2208

nsd_zones_primary 1
nsd_zones_secondary 0

nsd_zonestats_examplezone_queries_total 48
nsd_zonestats_examplezone_queries_by_type_total{type="A"} 41

[mozzieongit]

1. The elapsed time (`nsd_time_elapsed_seconds`) doesn't have much value
   currently, and doesn't reset when `nsd-control stats` is used, so I would
   probably remove it. However, it could be beneficial to reset this time
   whenever `nsd-control` resets the stats.

I disabled the elapsed time metric, until I point it to the remote-control time value.

2. For the zonestats block: Should the label name "zone" maybe renamed to
   "zonegroup" or something else?

I used zonestats, to copy the name from the config file.

3. Currently, the metrics don't have any HELP text (see [Prometheus Exposition Formats](https://prometheus.io/docs/instrumenting/exposition_formats/#comments-help-text-and-type-information))

Help and type annotations added.

[wcawijngaards]

The code looks good, and in general I like it, but I would want to change the copy of the statistics management code from remote.c into metrics.c, instead expose one or more static routines from remote.c and call them from metrics.c, making the code more understandable and maintainable.

[wcawijngaards]

The metrics options are not in nsd-checkconf.c, in the functions config_print_zone and config_test_print_server the options also need to be handled.

[wcawijngaards]

Can this have for example an implementation with if(clear) xfrd->nsd->rc->stats_time = *now;, similar to remote.c:3230. The comment could also be about something else.

[wcawijngaards]

The routines are copies of code at remote.c, I do not think that copying code that manages the inter-process communication part of the statistics like that is a very good idea. It would be better to make the routine public in remote.c and add its definition to remote.h and then call it here. The process_stats() function then needs to be changed to allow the different print_stats call than the print_stats call from remote.c.

[wcawijngaards]

Perhaps the path name extent is more precise with a quoted string.

Suggested change

	The HTTP path to expose the metrics at. Default is /metrics.
	The HTTP path to expose the metrics at. Default is "/metrics".