Drop node-exporter systemd metrics for units we do not care about

Some of these systemd metrics have large cardinalities, so it's good
to prune them at collection time.

Part of optimization work for issue #51.

roles/prometheus/templates/prometheus.yml.j2

-{# Common relabel config snippet #}
-{% macro relabel_configs(service=None) %}
-    relabel_configs:
-      - source_labels: [__address__]
-        regex: "([^.]*).*:[0-9]+"
-        target_label: host
-        replacement: "${1}"
-{% endmacro %}
-
 {# Generate static targets for hosts in an Ansible group #}
 {% macro targets_for_group(group, port) %}
       - targets:
@@ -80,12 +71,18 @@ alerting:
 {{ targets_for_service('prometheus', 'prometheus', 9093) }}
 
 scrape_configs:
-{{ job_static_config('node', 'all', 9100) }}
 {{ job_static_config('cgroups-exporter', 'all', 3909) }}
 {{ job_static_config('mtail', 'all', 3903) }}
 {{ job_static_config('docker', 'all', 9323) }}
 {{ job_static_config('backup-agent', 'all', 5331) }}
 
+{# The configuration for the node exporter drops some metrics with large cardinality. #}
+{{ job_static_config('node', 'all', 9100) }}
+    metric_relabel_configs:
+      - source_labels: [name]
+        regex: ".*\\.(path|scope|slice|socket|target|timer)"
+        action: drop
+
 
 {# Iterate over monitoring endpoints of all services. #}
 {% for service_name, service in services.items()|sort %}