feat(norms): Steward learns each metric's normal per hour of day, then week (#5739)

A metric registered with learn_normal() gets, for every resource reporting
it, a usual hold (the 90th percentile across past days of each hour's p95)
and a usual peak for each slot:
- an hour of the day once 5 days of history exist;
- an hour of the week once history reaches back 3 weeks, so weekly jobs are
  learned too.
Each slot takes in the hours either side, so a job that runs a little late
stays normal. One SQL statement relearns everything hourly into
metric_norms, from the rollup plus the raw window not rolled up yet;
norms_at() reads the slot a moment falls in. Nothing is registered yet:
Traefik (#5741) and hosts (#5742) opt in.

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

steward/backup/contents.py

@@ -30,6 +30,8 @@ KEPT: frozenset[str] = frozenset({
 # Rows that are measurements, logs or in-flight work: left out.
 EXCLUDED: frozenset[str] = frozenset({
     "plugin_metrics", "plugin_metric_latest", "plugin_metrics_hourly",
+    # Learned from the metrics above, so relearned once they fill again.
+    "metric_norms",
     "monitor_results",
     "email_outbox", "email_outbox_events", "password_resets",
     # Run history; its logs live on disk, not in the database.