feat(rid): screen one-hour spikes; refetch and redo-from to repair the store

RID's report sometimes carries a wrong value for one hour and corrects it
later. P.1 read 10.6 m at 2026-10-25 00:00 between 1.04 and 1.07 (RID now
says 1.06); the store kept it, because a report is re-fetched only while it
has holes, and the model issued a 10.8 m forecast for that hour.

- rid::judge / rid::screen: a reading that leaves the previous one (within
  3 h) by more than 1 m (2 m on the flashy tributaries P.76, P.77, P.81,
  P.82, P.84, P.85, P.87) while the next reading comes back near the
  previous level is a spike: not stored, a stored copy is deleted. Limbs
  and steps the next reading confirms are kept; the newest reading waits
  for the next one when it jumps more than twice the limit. On 2018-2026
  history: 16 spikes found (slipped decimals, sign flips, -5/-6/-9 m
  placeholders), no genuine change rejected.
- Every store goes through it (rid::store: run_cycle, fill-gaps, refetch).
- `ping-collector refetch FROM [TO]`: re-fetch reports regardless of holes
  (corrects revised readings, fills the discharge of level-only evening
  hours, removes stored spikes) and print the catch-up command when
  readings changed.
- `catch-up-forecasts --redo-from T`: delete the model forecasts issued
  from T on, then recompute them.
- Tests: judge/screen unit tests; a DB test of refetch, the live hold and
  the forecast deletion. Checked end to end against RID on a scratch DB.
- Docs: DATA_SOURCES (spikes, level-only hours), FLOOD_FORECASTING,
  rust/README.

docs/FLOOD_FORECASTING.md

@@ -643,7 +643,10 @@ Open-Meteo's estimate of rain that has since fallen, as in training, where
 the forward rain is observed rain. Such rows are written only where no model
 row exists; `created_at` tells them apart (more than an hour after `as_of`).
 `ping-collector catch-up-forecasts --hours N` does the same further back (up
-to 60 days).
+to 60 days); with `--redo-from T` it first deletes the model forecasts issued
+from T on, to recompute them after `ping-collector refetch` corrected the
+readings they came from (a bad reading reaches the features for 72 h through
+`roll_max_72`).
 
 A model row in `forecast_history` is never replaced by a heuristic one. While
 RID is silent overnight each hourly run re-issues the 21:00 reading, and once