fix(insights): correlate residuals, not the calendar

Two daily series that both follow the season (temperature and blood
pressure, daylight and sleep) correlated strongly with nothing between
them. In a seeded simulation of such a pair the engine surfaced a
finding in 58 % of 400-day runs and 30 % of 180-day runs, and a real
next-day effect under the season came back with the wrong sign.

Every daily pair now goes through seasonallyAdjustedPearson: a linear
trend and one annual harmonic are removed from both series (trend only
under 120 paired days), the residuals are correlated, and the p-value
uses the Pyper-Peterman effective sample size. Shrinkage and the
phrasing tier read the effective size too. BH-FDR and the effect floor
are unchanged. Pairs too short to fit the basis are not tested.

The weather channels are tested once over the mean of the day before
and the day itself against the same day's outcome (lagDays 0), which
pairs the overnight low with the night it shaped now that sleep is
keyed on the wake day. Discovered pairs carry an additive nEff, the
same-day templates are wired, and the matrix skips the weather read
when the environment module is off.

scripts/compare-correlation-findings.ts counts findings per account
before and after, read-only, for the release-candidate dry run.

docs/api/openapi.yaml

@@ -34740,6 +34740,9 @@ components:
           type: integer
           minimum: -9007199254740991
           maximum: 9007199254740991
+        nEff:
+          description: v1.42 — effective sample size after serial correlation, ≤ `n`; the p-value uses it.
+          type: number
         r:
           type: number
         pValue:
@@ -62417,13 +62420,18 @@ components:
           type: integer
           minimum: -9007199254740991
           maximum: 9007199254740991
-          description: Paired-day count after the day+1 lag join (≥ 20).
+          description: Paired-day count after the lag join (≥ 20).
+        nEff:
+          description: v1.42 — effective sample size after serial correlation (Pyper–Peterman), ≤ `n`. The p-value, the shrinkage
+            and the tier use it.
+          type: number
         r:
           type: number
-          description: Pearson r over the lag-joined daily series.
+          description: "Pearson r over the lag-joined daily series, computed on residuals: a linear trend (and, from 120 paired
+            days, one annual harmonic) is removed from both series first (v1.42)."
         pValue:
           type: number
-          description: Two-sided exact Student-t p-value (< 0.05).
+          description: Two-sided exact Student-t p-value (< 0.05) at the effective sample size.
         qValue:
           type: number
           description: Benjamini-Hochberg FDR-adjusted q-value (≤ the surface threshold).
@@ -62446,7 +62454,8 @@ components:
           type: integer
           minimum: -9007199254740991
           maximum: 9007199254740991
-          description: Lag in days applied (1).
+          description: "Lag in days applied: 1 = behaviour day, outcome the next day; 0 = the environmental channels, whose
+            exposure is the mean of the day before and the day itself, paired with the same day's outcome (v1.42)."
         window:
           description: v1.22 — which window surfaced the pair. Absent on the 180-day scan (retrospective default); `recent` for an
             emerging early-detection pair.