find_node_context_host calls can be more expensive as the balancer increases in size. That led to particular scaling problems with internal_find_best_byrequests in MODCLUSTER-757, which repeated a context_host_ok/find_node_context_host for every balancer member. So we should likewise consider avoiding any other context_host_ok/find_node_context_host call repeated in a request for more optimal performance.

It looks like the following paths can allow for a few repeat calls for each request, though these wouldn't scale up in call counts with balancer members to be as large a concern as with internal_find_best_byrequests in MODCLUSTER-757:

1. proxy_cluster_pre_request->find_session_route->find_route_worker->context_host_ok/find_node_context_host
2. proxy_cluster_trans->get_route_balancer->find_node_context_host
3. proxy_cluster_trans->get_context_host_balancer->find_node_context_host