feat: NIXL Telemetry to be deployed with docker compose (#7676)

Signed-off-by: Aleksandr Bilkovskii <alexanderb@nvidia.com>

deploy/observability/prometheus.yml

@@ -48,6 +48,14 @@ scrape_configs:
     static_configs:
       - targets: ['host.docker.internal:8081']
 
+  # NIXL Prometheus metrics (requires env on the worker: NIXL_TELEMETRY_ENABLE=y,
+  # NIXL_TELEMETRY_EXPORTER=prometheus, NIXL_TELEMETRY_PROMETHEUS_PORT=19090).
+  # Compose Prometheus reaches the host via host.docker.internal (see docker-observability.yml).
+  - job_name: 'dynamo-nixl'
+    scrape_interval: 1s
+    static_configs:
+      - targets: ['host.docker.internal:19090']
+
 
   # KVBM leader related metrics
   - job_name: 'kvbm-metrics'