Add metrics relation

[rgildein]

Expose metrics port and provide metrics-endpoint relation. With these changes, charm can be monitored with Prometheus.

How it can be tested:

$ tox -e integration -- --model kubeflow --keep-models
$ juju exec --unit grafana-agent-k8s/0 -- curl localhost:12345/agent/api/v1/metrics/targets | jq '.data.[] | select(.labels.juju_charm == "kserve-controller") | .endpoint'
# using endpoint obtained from previous command
$ juju exec --unit grafana-agent-k8s/0 -- curl http://10.1.23.229:8080/metrics
# HELP certwatcher_read_certificate_errors_total Total number of certificate read errors
# TYPE certwatcher_read_certificate_errors_total counter
certwatcher_read_certificate_errors_total 0
# HELP certwatcher_read_certificate_total Total number of certificate reads
# TYPE certwatcher_read_certificate_total counter
certwatcher_read_certificate_total 1
...
workqueue_work_duration_seconds_count{name="inferenceservice"} 385
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="1e-08"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="1e-07"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="1e-06"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="9.999999999999999e-06"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="9.999999999999999e-05"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="0.001"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="0.01"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="0.1"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="1"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="10"} 0
workqueue_work_duration_seconds_bucket{name="trainedmodel",le="+Inf"} 0
workqueue_work_duration_seconds_sum{name="trainedmodel"} 0
workqueue_work_duration_seconds_count{name="trainedmodel"} 0

[orfeas-k]

Good job @rgildein !