SPIKE: Shard some traffic to us-east-2 read replica

This ticket tracks determining what it would take to shard, say, 5% of quay.io read traffic to the us-east-2 read replica.

While sharding traffic would like increase resilience and facilitate disaster recovery, It is suspected/known that sharding would increase financial costs and latency. We should try to quantify those costs, so that a proper cost/benefit analysis could be done.

It would also be good to evaluate what would need to change architecturally in the quay code to support selective direction of traffic to read-replicas.