Enable scraping of application-level metrics on `dev.gitlab.org`

Currently the dev.gitlab.org runs in Azure (it seems), and we do not scrape the dev.gitlab.org application metrics from Unicorn/Puma. The /-/metrics endpoint.

This makes it impossible to understand exact implications of switching from Unicorn to Puma, and to Puma in lower amount of workers, as we miss detailed application level details about:

  1. GC behavior,
  2. Worker restarts due to over-memory-threshold,
  3. Timeouts on requests, etc.
Edited by Kamil Trzciński