We're currently using 2 connection pools when connecting to the
Redis cache instance. This makes Rails.cache and Gitlab::Redis::Cache
use the same pool so that we can reuse connections more efficiently.