Increase Patroni's patience when talking with Consul

Increase Patroni's DCS settings for retry_timeout (currently 10 seconds) and ttl (currently 30 seconds).

The rationale is described here: https://gitlab.com/gitlab-com/gl-infra/infrastructure/issues/7790#note_213228185

And remember that ttl is silently halved by Patroni before being sent to Consul.

The ttl should probably be several times as large as the retry_timeout.

Note: The loop_wait setting (currently 10 seconds) may be ok to leave as-is, even though we are increasing the other settings. loop_wait is roughly how long Patroni waits passively for notification from Consul of a cluster state change before it starts another round of pushing its own status updates to Consul and renewing its Consul session TTL. The configured ttl must be significantly larger than both loop_wait and retry_timeout.