This patch replaces the plain pod with with a more sustainable
deployments which will also survive draining and automatic recreation of
pods.

Reference:
https://gitlab.com/gitlab-org/ci-cd/gitlab-runner-pod-cleanup

This patch should provide an automated backup for all postgresql
clusters in the cluster. This implementation is less than ideal, due to
technical sharing of backups among all namespaces, but for now, it's
better than no backup. There will be further work on separating these
backups, but it is what it is for now.

This patch fixes the current issue with the network policy regarding
host network containers, calico and Kubernetes NetworkPolicies.

The problem originates from the selected bind port to reach containers
on the calico overlay network, which is using the wireguard endpoint,
instead of the host external IP. Since these also always change, and are
part of the Pod-CIDR, it's not really possible to selected them using an
ipBlock without opening things up for all pods.

The reason why the Kubernetes NetworkPolicies don't work, is due to them
not being applied to host network pods, therefore the control-plane
can't be really selected here. therefore calico network policies with
automatic host endpoints are a requirement.

This patch adds two services, one for web, which can be exposed by an
ingress, and one of type LoadBalancer that is exposed directly, to cut
down on transfer overhead using a relay or alike.