Fix juju_unit and juju_application labels

[Abuelodelanada]

Issue

This PR fixes #134

Solution

Based on the PR created by @dashmage, now we are able to properly add relabel_configs for juju_application and juju_unit

Testing Instructions

In an lxd model deploy the following:

default-base: [email protected]/stable
saas:
  prometheus:
    url: microk8s:admin/cos.prometheus
applications:
  cos-proxy:
    charm: local:cos-proxy-3
    num_units: 1
    to:
    - "10"
    constraints: arch=amd64
  nova-compute-kvm:
    charm: ubuntu
    channel: latest/stable
    revision: 24
    num_units: 1
    to:
    - "6"
    constraints: arch=amd64
    storage:
      block: loop,100M
      files: rootfs,100M
  nrpe:
    charm: nrpe
    channel: latest/edge
    revision: 122
    options:
      nagios_host_context: bootstack-blue
machines:
  "6":
    constraints: arch=amd64
  "10":
    constraints: arch=amd64
relations:
- - nrpe:general-info
  - nova-compute-kvm:juju-info
- - cos-proxy:monitors
  - nrpe:monitors
- - cos-proxy:downstream-prometheus-scrape
  - prometheus:metrics-endpoint

In a microk8s model deploy cos-lite (only prometheus and traefik are needed for the test though) using the offers overlay:

juju deploy cos-lite --channel=edge --trust --overlay ./offers-overlay.yaml

Offer Prometheus' metrics-endpoint interface so avoid installing g-agent:

juju offer cos.prometheus:metrics-endpoint 

Configure nagios_host_context:

juju config nrpe nagios_host_context="bootstack-blue"

Relate cos-proxy to prometheus:

juju relate cos-proxy prometheus    

You'll end up wit a deployment like this one:

Once every relation is in place, we can see in Prometheus that juju_application and juju_unit are ok:

[lucabello]

LGTM

We should probably revert the fetch-lib of cos_agent.py because the pydantic pin will fail the CI :(