perf(metrics): index plugin_metrics for host/dashboard reads

plugin_metrics had only a PK on id, so every host-detail, full-metrics, and
dashboard-widget load sequentially scanned the entire time-series table — which
grows by (sources × resources × sample cadence), so the host views got slower
over time (operator-reported "blocked/slow" loads). monitor_results was already
indexed, so the uptime aggregation wasn't the bottleneck.

Add two composite indexes matching the hot query shapes:
- (source_module, resource_name, recorded_at) — history range scans, fleet/
  widget queries, and the 'host:%' sub-resource prefix.
- (source_module, resource_name, metric_name, recorded_at) — the latest-value-
  per-metric group-by/self-join.

Model __table_args__ + core migration 0023. Integration lane validates the
migration via `alembic upgrade head`.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016Jg27rgypiW2efULXJDtMC

steward/migrations/versions/0023_plugin_metrics_indexes.py

@@ -0,0 +1,39 @@
+"""Index plugin_metrics for host/dashboard reads
+
+plugin_metrics had only a PK on id, so every host-detail / full-metrics /
+dashboard-widget query (all filter by source_module + resource_name over a
+recorded_at range) sequentially scanned the whole time-series table — slower as
+samples accumulate. Add the two composite indexes that match those query shapes.
+
+The non-concurrent CREATE INDEX takes a brief exclusive lock; it runs once at
+startup migration time, acceptable for this table.
+
+Revision ID: 0023_plugin_metrics_indexes
+Revises: 0022_unify_monitors
+Create Date: 2026-06-20
+"""
+from typing import Sequence, Union
+from alembic import op
+
+revision: str = "0023_plugin_metrics_indexes"
+down_revision: Union[str, None] = "0022_unify_monitors"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_index(
+        "ix_plugin_metrics_module_resource_recorded",
+        "plugin_metrics",
+        ["source_module", "resource_name", "recorded_at"],
+    )
+    op.create_index(
+        "ix_plugin_metrics_module_resource_metric_recorded",
+        "plugin_metrics",
+        ["source_module", "resource_name", "metric_name", "recorded_at"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_plugin_metrics_module_resource_metric_recorded", "plugin_metrics")
+    op.drop_index("ix_plugin_metrics_module_resource_recorded", "plugin_metrics")

steward/models/metrics.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 import uuid
 from datetime import datetime, timezone
-from sqlalchemy import DateTime, Float, String
+from sqlalchemy import DateTime, Float, Index, String
 from sqlalchemy.orm import Mapped, mapped_column
 from .base import Base
 
@@ -17,3 +17,14 @@ class PluginMetric(Base):
     recorded_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), nullable=False, default=lambda: datetime.now(timezone.utc)
     )
+
+    # This time-series table grows by (sources × resources × sample cadence); every
+    # host-detail / full-metrics / dashboard-widget read filters by
+    # (source_module, resource_name) over a recorded_at range. Without these it's a
+    # full sequential scan on every load.
+    __table_args__ = (
+        Index("ix_plugin_metrics_module_resource_recorded",
+              "source_module", "resource_name", "recorded_at"),
+        Index("ix_plugin_metrics_module_resource_metric_recorded",
+              "source_module", "resource_name", "metric_name", "recorded_at"),
+    )