Update metrics endpoint relation data every re-init

[sed-i]

Issue

• External hostname is not updated if an ingress is added *after* relating a charm to Prometheus #368: metrics endpoint is not updated to ingressed address after relating to traefik.
• Having the lib observe update-status by default would probably solve the issue, but user would have to wait for the update-status hook interval to elapse.
• Instructing the user to pass multiple refresh events to the MetricsEndpointProvider constructor would work only after ready_for_unit is emitted too early traefik-k8s-operator#78 is fixed, but also introduces additional friction for the user to remember to pass custom refresh events.

Solution

Call self._set_scrape_job_spec in __init__ to update app data every charm re-init.

Fixes #368.

Drawbacks

Adds unnecessary comms with juju.

Context

• This is an alternative to Always refresh metrics endpoint reldata on update-status #385.
• May be related to loki ingress endpoint doesn't work loki-k8s-operator#202.

Testing Instructions

Release Notes