fix(grafana): move or remove the last 6 UI-made alert rules (#1690)

[groupsky]

Closes #1690.

Owner decision from #1690: remove 2 UI-made rules that a provisioned rule covers, and move 4 into the repo. After this, folder "General Alerting" has no UI-made rules.

Heat pump overload: the provisioned rule did not cover the UI rule

The issue asked to confirm this before deleting xjXziW_4k. It did not match:

	UI rule xjXziW_4k	provisioned webhook-lA5nSTt4k (before)
Query	mean(p) of current_power, heat_pump, grouped by phase	max(p), all phases in one series
Condition	avg over 15 min > 1000 W on any phase	avg over 10 min > 3000 W
for	15m	10m

So the provisioned rule needed one phase above 3000 W. Data on routy:

• 2026-01-10..24 (winter): phase B was above 1000 W (15 min mean) in 42 of 1344 windows. The highest reading on any phase was 1956 W, so the old provisioned rule could never fire.
• The UI rule really fired 23 times since 2023-11 (last 2026-04-09), from the Grafana state history.
• Last 7 days: highest 15 min mean per phase is 658 W, so the change does not fire now.

Decision (option A, agreed): webhook-lA5nSTt4k now has the UI rule's meaning: per-phase mean > 1000 W over 15 min, for 15m. The message says "над 1000 W на поне една фаза ... (3 x 1000 W = 3000 W общо)". Then xjXziW_4k is deleted.

Other rules

UI uid	What happens
3057STtVkz AC alert (210/248 V, no for)	deleted; provisioned webhook-3057STtVkz (207/253 V, 4m) stays
RJcnSop4z External alert	moved to webhook-RJcnSop4z in heat-pump-alerts.yaml: min(s2Temp) of monitoring2 heatpump-ctrl, avg < 5 °C, for 5m
eAc7ITtVk Output alert	moved to webhook-eAc7ITtVk: max(s3Temp), avg > 60 °C, for 5m
rA5nSotVkz Temperature alert	moved to webhook-rA5nSotVkz in new controlbox-alerts.yaml: mean(t) of xymd1 controlbox (t < 80), outside 5-37 °C, for 5m. The message now says 5-37 (it said 5-35).
j05nSotVz Humidity alert	moved to webhook-j05nSotVz: mean(h), outside 15-85 %, for 5m

The 4 moved rules use the exact query, reducer, window and for read from the Grafana DB copy. noDataState and execErrState stay OK as in the UI rules, so a short InfluxDB restart does not page (#1682). Deletes go through deleteRules in delete-legacy-duplicate-alerts.yaml, like #1683.

Deploy without a page

On routy at 15:04 UTC, all 6 UI rules and webhook-lA5nSTt4k are Normal, so the new copies start Normal and do not page.

Tests

• validate-grafana-alerts: passes for the 3 changed files.

[groupsky]

Review and local test

Local Grafana 9.5.21 + InfluxDB 1.12.4 (rule interval 10s and for: 0s only for speed):

• normal data (all phases 500 W, s2Temp 10, s3Temp 40, t 20, h 50): all 5 rules Normal.
• bad data (phase B 1500 W, A/C 500 W, s2Temp 2, s3Temp 70, t 40, h 90): all 5 rules Alerting. So the overload rule fires when one phase is above 1000 W.
• t = 100 only: controlbox temperature stays Normal (the t < 80 filter drops sensor errors).
• All rules health ok, no evaluator or provisioning errors.

Subagent review: no blockers. It checked each moved rule field by field against the Grafana DB export. Fixed: the overload message now says plainly that the limit is per phase. Kept as before: the overload rule keeps noDataState: NoData (same as the old provisioned rule and the UI rule; a short restart is shorter than its 15 min window).

Before deploy I check again that all 6 UI rules and webhook-lA5nSTt4k are Normal, so no deleted rule sends a "resolved" message.