Implement `geo_secondary_role` to automate some configurations

Description

This is a follow up on the work to enable Geo PostgreSQL service for EE: #1977 (closed). Following the discussion in slack: https://gitlab.slack.com/archives/f_disaster-recovery/p1489077610571850 with the "Role pattern" we already used for sentinel, by enabling the role we can trigger other configurations automatically and make easier for the end-user to enable the geo_secondary "feature"

Proposal

When geo_secondary_role is enabled, we should:

• Enable geo_postgresql
• Fill in database.yml for geo secondary tracking database with geo_postgresql data
• Enable variables to enable replication of regular database as described here: https://docs.gitlab.com/ee/gitlab-geo/database.html#step-2-configure-the-secondary-server

cc @stanhu @marin @dbalexandre @regisF @rspeicher