Add support for read replicas

[wslulciuc]

In the Jdbi docs, it states that if we want to have two separate pools (reads / writes) we need to manage separate Jdbi instances:

If you want to have two separate pools, for example a read-only set that connects to read replicas and a smaller pool of writers that go only to a single host, you currently should have separate Jdbi instances each pointed at a separate DataSource.

Let's add support to Marquez to allow reads from read replicas.