stuck in maintenance after no-op upgrade

[Guillermo Gonzalez]

The charm gets stuck in maintenance status, only when it's related to postgresql *and* prometheus.

Tracked things to a bad reactive state mix, the .unit-state.db ends up with: `reactive.states.prometheus-client.relation.configured` set and `plugins.prometheus-client.configured` unset, this causes it to never execute the configure handler (configure_telegraf_deb or configure_telegraf_snap)

Maybe it's something else, but deleting the `reactive.states.prometheus-client.relation.configured` state from the DB and re-running any hook (update-status) fix things.

Steps to reproduce (tested on xenial)

juju deploy cs:postgresql
juju deploy cs:prometheus2
juju deploy cs:telegraf
juju add-relation telegraf postgresql:juju-info
juju add-relation telegraf postgresql:db
juju add-relation telegraf prometheus2:juju-info
juju add-relation telegraf prometheus2:target
charm pull cs:telegraf
juju upgrade-charm telegraf --path ./telegraf