Improve performance of metrics collectors

[GeorgeTsagk]

Description

We currently have a couple of different collectors that are working every time a prometheus scrape occurs.
The default timeout for these scrapes is 10s, and under loaded environments some of the queries performed by those collectors may introduce great delays, possibly extending the metric collection beyond the 10s default timeout.

Proposed solution

Identify the resource-intensive data collections and have them running independently in the background. This way the "heavy" calculation will be avoided at scrape time, and will instead be replaced by a system that's instantly providing the latest metric values.