Improve HA documentation based on latest requirements of running application

We should create and update HA documentation to take into account the following:

1. Unicorn zero-downtime is not longer supported
2. We require at least two nodes for HA and zero-downtime deployment
3. The nodes needs to be doing rolling update
4. The load balancer should use /-/readiness endpoint to detect node not accepting traffic
5. Document behaviour of blackout period during service restart

This is needed for safe rollout of Puma and Puma switchover as

• Puma requires full service restart, unlike Unicorn continuing to process traffic while restarting,
• Unicorn and Puma supports blackout period: non-interrupting way to restart service, by disconnecting it from load balancer
• For HA it requires that at least two nodes to be running, otherwise all nodes would be disconnected from the LB,