Summary

Previous dual read matches URI properties only when the versions are the same (same ZK data served from ZK vs Observer), but with Kafka data served, the version will never be the same. And the OutOfSync metric was never useful since only one version of uri properties is cached for each cluster, when uri data changes frequently, the OutOfSync increases and never comes down.
This new URI matching now produces a similarity metric which represents the fraction of matching uris present in the ZK response and the observer response, just the same as how observer is monitoring the cluster uri similarity. The similarity metric may go up during downstream deployment/updates, but eventually it should be 1. The goal is to verify that client side response processing logic in INDIS is not causing uri data to drift (given that Kafka data matches ZK data on observer already).

The Cluster and Service monitors remain unchanged.
(This new monitor should be as expensive as the previous one since the previous one called the .equals method on UriProperties, which compares every single URI).

Test Done

Added unit tests. Will do manual qei testing with d2-proxy in the counterpart container PR for adding the config.