Adds metrics to detect routing and proxy issues:

Router:
- decisions_total: local vs remote routing split by target region
- errors_total: no_running_instances, no_sentinels_for_instances, config_not_found
- duration_seconds: route decision latency (cache + DB)
- running_instances: distribution of available instances per decision

Proxy:
- forward_total: success/failure by target and error type (timeout, conn_refused, etc.)
- backend_duration_seconds: backend response latency by target
- backend_response_total: HTTP status classes (2xx/4xx/5xx) from backends
- hops: cross-region hop count distribution