Add test for DCS failsafe mode, simulating k8s API being down

https://github.com/canonical/postgresql-k8s-operator/pull/677 enables DCS failsafe mode which should improve stability in overloaded environments where the k8s can become temporarily unavailable.

We should add integration tests for that scenario, stopping or otherwise disabling access to the k8s API and verify that the primary doesn't step down. We should also verify that the primary will step down if it cannot connect to all the replicas.

Start by reading PR #677 and the repository's existing integration tests to find how DCS failsafe mode and Kubernetes API access are exercised. Add coverage for a temporarily unavailable Kubernetes API, verifying that the primary stays up, and for failure to connect to all replicas, verifying that the primary steps down.

Written by the indexing model from the issue text.