PostgresqlNotEnoughConnections fires as a false positive on small deployments

The alert fires constantly on a lightly-loaded deployment even though PostgreSQL is healthy and the application is working correctly.

## Steps to reproduce

1. Deploy postgresql-k8s and relate to cos (or cos-lite)
2. Relate postgresql-k8s to a consumer charm and don't use it very much (I am using [joplin-server](https://charmhub.io/joplin-server))
3. Check the firing alerts

## Expected behaviour
No alerts when the deployment is healthy

## Actual behaviour
Continuous alerts

## Versions

Operating system: Nixos 25.11

Juju CLI: 3.6.20-genericlinux-amd64

Juju agent: 3.6.14

Charm revision: 774

microk8s: -
k3s: v1.34.5+k3s1

## Log output

Juju debug log:

-

## Additional context

-

Start by reproducing the deployment with postgresql-k8s, cos or cos-lite, and a lightly used consumer such as joplin-server, then inspect the PostgresqlNotEnoughConnections alert condition. Compare its behavior with the reported healthy, low-traffic deployment; done means the alert no longer fires continuously when PostgreSQL is healthy.

Written by the indexing model from the issue text.