Make gitlab_runner_jobs_total metric to be present always

The gitlab_runner_jobs_total metric is not present after the process is started. We had a similar problem with few other metrics (like gitlab_runner_jobs) in the past. That may make some things to fail.

For example, in GitLab.com monitoring we are looking at this metric, compare it to a previous reading from an hour ago, and make an alerting decision based on that. If the metric disappears because of the runner process restart (which must happen when we, for example, update runner's version), then the alert is triggered creating a false feeling of an ongoing problem.

Making the metric present always, being initiated as 0 at the process start (as it does for gitlab_runner_jobs) should solve this problem.