fix: use external_url in alert source

[lucabello]

Issue

Closes #581.
Closes canonical/alertmanager-k8s-operator#206.

Not sure why we were using the route_prefix from the self.external_url (from ingress), but adding it to the self.internal_url (which is always the FQDN).

Solution

Use the external URL instead of the internal one.

Screenshots

Before:

After:

Testing Instructions

juju deploy alertmanager-k8s --channel=latest/edge alertmanager
juju deploy prometheus-k8s --channel=latest/edge prometheus
juju deploy traefik-k8s --channel=latest/edge traefik

juju relate prometheus:alertmanager alertmanager
juju relate prometheus:metrics-endpoint alertmanager
# observe the watchdog alert's source link (internal url)
juju relate prometheus:ingress traefik
# wait for things to settle
# observe the same alert's source link again and try to click it!

Upgrade Notes

Anything that refreshes the Pebble plan will activate this change, so the upgrade should be seamless.

[PietroPasotti]

cough tests cough

[lucabello]

A unit test has been fixed :)

I also skipped the tests we were XFail-ing, because they'll just waste time in integration tests without contributing :)

addressed

[PietroPasotti]

In principle the change is OK and given it's high prio I don't want to block it, but we're skipping a huge amount of tests for unclear reasons.
Can you explain what's going on and remove any tests that are no longer necessary?