Allow additional method of gathering web metrics

Today in heavily tested applications that are loaded with stress tests, the /metrics endpoint might not be available or is available with very high latency. Prometheus has very strict scrape_timeout that often results in missing metrics and in such cases, we are unable to gather metrics from that period. Exposing metrics on a separate endpoint and running on master side-steps that problem.

Proposal

Introduce:

gitlab:
  monitoring:
    web_exporter:
    #  enabled: true
    #  address: localhost
    #  port: 3808