test(alerts): seed metrics the way ingest does, through record_metrics (#5473)

#5473 "Opening an alert rule takes up to 15 s". With the rule pickers now
reading plugin_metric_latest, nine integration tests failed on rule_fields:
"host_agent has no resource 'box-…'". They seeded a resource by adding a raw
PluginMetric row, which no production path does: host-agent ingest and
record_metric both go through record_metrics, which keeps the latest-value
table. The seeds now go through record_metrics too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/integration/test_host_agent_alerts.py

@@ -31,6 +31,7 @@ def test_ingest_fires_a_host_rule_from_a_backlog(app):
     from steward.alerts import service as alert_svc
     from steward.models.alerts import AlertEvent, AlertState, AlertStateEnum
     from steward.models.hosts import Host
+    from steward.core.alerts import record_metrics
     from steward.models.metrics import PluginMetric
     from steward.models.users import User, UserRole
 
@@ -50,8 +51,8 @@ def test_ingest_fires_a_host_rule_from_a_backlog(app):
                     "VALUES (:id, :host, :hash, now(), now(), now())"),
                     {"id": str(uuid.uuid4()), "host": host.id,
                      "hash": hashlib.sha256(token.encode()).hexdigest()})
-                s.add(PluginMetric(source_module="host_agent", resource_name=hostname,
-                                   metric_name="cpu_pct", value=5.0))
+                await record_metrics(s, [PluginMetric(source_module="host_agent", resource_name=hostname,
+                                                      metric_name="cpu_pct", value=5.0)])
                 await s.flush()
                 fields = await alert_svc.rule_fields(
                     s, name=f"cpu {tag}", source_module="host_agent", resource_name=hostname,