archgw: address drift in prometheus cluster name

[codefromthecrypt]

Recent changes made some config incompatible and fixed the model name metrics glitch by no longer recording the model name.

$ curl -sl http://localhost:19901/stats?format=prometheus|grep ollama|grep -v ' 0' |tail -2
envoy_cluster_docker_internal_upstream_rq_time_sum{envoy_cluster_name="ollama_host"} 1455
envoy_cluster_docker_internal_upstream_rq_time_count{envoy_cluster_name="ollama_host"} 2