pgbouncer-leader-check.service can stop indefinately in times of failure

the systemd unit pgbouncer-leader-check.service which we use on the pgbouncer as health checks for the GCP loadbalancer follows the standard systemd behaviour when it fails. This means systemd will attempt to restart it and if it crash loops too much, systemd will just give up entirely.

This means in times of failure, it's possible that if problems occur long enough, the unit will never get restarted (after crash looping too much).

We need to update this systemd unit so that systemd will be less aggressive restarting it, and always attempt to restart it indefinitely