Add collector for systemd service status.

[alienth]

Utilizes dbus API to determine state of services managed by systemd. It will track the state of services listed under [[SystemdService]] entries in the config (modeled off of processes_windows collector).

It writes metric information to linux.systemd.unit, and to os.service.running. The systemd metric tracks each possible ActiveState of a systemd service unit. The os metric simply tracks whether the service is running or not.

If WatchProc was set to true in the config, it also attempts to lookup the service process and sends process metrics via the processes_linux collector methods.

Example config:

[[SystemdService]]
   Name = "scollector"
   WatchProc = true

[[SystemdService]]
   Name = "redis-.*"

[gbrayut]

This file is missing a "bosun.org/util" import. Adding that should fix travis ci builds

[captncraig]

I made a new way to do this kind of init stuff, but unfortunately didn't convert all these old ones in main. See what I did to the haproxy one here: https://github.com/bosun-monitor/bosun/blob/master/cmd/scollector/collectors/haproxy_unix.go#L16

That should make the collectors more self contained.

[gbrayut]

I would add osServiceRunningDesc = "0: active, 1: inactive" as a const in scollector/collectors/collecors.go and use that for the os.service.running description. That way it can be shared with windows. You can also move the metric name to a const in that same file (ex: osServiceRunning = "os.service.running")

[alienth]

@captncraig / @gbrayut Requested changes made.

I pushed the followup commits as-is to make review easier. I plan to squash things down before merge.

[gbrayut]

LGTM... will try and test it tomorrow.

[alienth]

Squished down. I left the NameMatches change separate because my changes no longer require it, but it should still be done.