Improve reboot-patroni-cluster Ansible Playbook, to forcefully close database client connections when draining connections from a node

Improve the https://ops.gitlab.net/gitlab-com/gl-infra/db-ops/-/blob/master/ansible/reboot-patroni-cluster/reboot_cluster.yml playbook to perform a clean shutdown of database clients by Pausing and shutting down node pgbouncers during the node draining (before reboot). As discussed at production#7712 (comment 1104659359)

Without this improvement the draining can take too long (around 40 minutes per node, which leads to several hours for the whole cluster).

Steps

First PAUSE pgbouncers:

for i in $(ls -1 /usr/local/bin/pgb-*); do $i -c "PAUSE"; done;

Then check the pool status, wait for sv_active to drop to 0:

for i in $(ls -1 /usr/local/bin/pgb-*); do $i -c "SHOW POOLS"; done;

Then SHUTDOWN pgbouncers:

for i in $(ls -1 /usr/local/bin/pgb-*); do $i -c "SHUTDOWN"; done;