* Refactor scale down logic to not force remove unreachable instances

* Add unit tests

* Address lint warnings

* Add sleep and idle timeout in test code in an attempt to stabilize tests

* Address lint warnings

* Add sleep after killing pod to give test environment time to stabilize

* Wait for unit to be removed from cluster after force removing it

* Wait for forcefully removed unit to actually be removed from the cluster

* Add integration test for recovery after a network cut

* Remove duplicate code introduced after merging main

* Install helm3 in CI runner for self healing integration tests

* Fix failing unit test

* Pin flake8 to v5.0.4 due to incompatibilities with v6

* Install helm on CI runners using sudo

* Port full cluster crash recovery logic from vm charm + add full cluster crash integration test

* Replace extending pebble restart delay with using on-failure and on-success

* Remove unused manifest file

* Fix call to datetime.datetime.now()

* Address PR feedback; recover cluster from full crash if one node

* Avoid running update status when the cluster is in waiting state (e.g. for tls or initialization)

* Attempt at bugfix for the full cluster crash recovery

* Fix broken lint and unit tests

* Fix lint warnings

* Fix typo used to test locally

* Fix minor bugs in full cluster crash recovery logic + make integration test pass

* Fix lint warnings

* Address PR feedback