Expand Prometheus metrics

[mgrabovsky]

As a follow-up to #419, it would be nice to expand the list of exposed metrics in the future. Some useful ones that come to mind are (roughly in order of descending priority):

• Task duration summaries (min, max, quantiles) over some reasonable window
• Number of tasks (failed, successful) per operating system release
• Backtrace rating histogram, i.e. the number of successful tasks that produced a backtrace with rating 0..4. (This would require calling abrt-action-analyze-backtrace on the result.)
• CPU load (1/5/15 min)
• Memory usage
• httpd stats
• Time until TLS certificate expires
• Occurrence of common retracing errors
• Server uptime
• Retrace Server version
• Uploaded coredump size summaries

It would also be nice to split some of the metrics by operating system (e.g. tasks failed for Fedora 34, tasks failed for CentOS 8 etc.).