fix: correct metric names in llama-swap monitoring (llamacpp_* -> llamaswap_*), update alerts + dashboard + scrape config

ansible/roles/llm-inference-multimodel/templates/llama-swap-prometheus-scrape.yml.j2

@@ -30,7 +30,10 @@
   metrics_path: /metrics
   
   # Relabeling: extract model name from metric labels for dashboard grouping
-  metric_relabel_configs:
-    - source_labels: [__name__]
-      regex: 'llamacpp_.*'
-      action: keep
+# llama-swap exposes llamaswap_* metrics (GPU VRAM, utilization, power, CPU,
+# network, load average). Per-model inference metrics are not available at the
+# proxy level. Filter to keep only llamaswap_* metrics to reduce cardinality.
+    metric_relabel_configs:
+      - source_labels: [__name__]
+        regex: 'llamaswap_.*'
+        action: keep