fix: Optimize Prometheus scrape target discovery

[hisarbalik]

Description

A MetricPipeline using the input prometheus on a cluster with an high amount of pods will result in a DaemonSet of the telemetry-agent where every instance will consume high memory, even if not data get processed.
Every single agent is processing only the pods located on the related node, however, the discovery mechanism does not respect the node sharding and will discover all pods. This problem gets fixed by this PR by adjusting the discovery to exclude already pods not located on the same node as the agent.

Changes proposed in this pull request (what was done and why):

• The metric agent should discover the scrape targets deployed on the same node as the agent instance is running, this will reduce memory and CPU consumption on larger clusters significant. Configuration documentation can be found here

Changes refer to particular issues, PRs or documents:

Traceability

• The PR is linked to a GitHub issue.
• The follow-up issues (if any) are linked in the Related Issues section.
• If the change is user-facing, the documentation has been adjusted.
• If a CRD is changed, the corresponding Busola ConfigMap has been adjusted.
• The feature is unit-tested.
• The feature is e2e-tested.