fix(vital_signs): use circular variance for wrapped phase values

[yahyasaqban-lab]

Summary

Phase values from atan2(i\_val, q\_val) are wrapped to (-\pi, \pi]). Linear variance treats values near (+\pi) and (-\pi) as (\sim 2\pi) apart when they are physically (\sim 0.003) rad apart, producing variance of (\sim \pi^2) instead of (\sim 10^{-6}).

Replace with circular variance (1 - R̄, where R̄ is the mean resultant length), which correctly handles wrapped circular data.

Repro

import math
phases = [math.pi - 0.001, -math.pi + 0.001]
# Linear: ~9.87  (broken)
m = sum(phases) / 2
print('linear:  ', sum((p - m) ** 2 for p in phases) / 2)
# Circular: ~5e-7  (correct)
s = sum(math.sin(p) for p in phases)
c = sum(math.cos(p) for p in phases)
n = len(phases)
print('circular:', 1 - (s*s + c*c)**0.5 / n)

Impact

Fixes the (\pm 15) bpm jitter in heart-rate detection and jumpy vitals reported in issues #485 and #519.

Fixes #593

[ruvnet]

Closing as a duplicate of #595, which was filed by the original reporter (@akhilesharora) of #593 on 2026-05-16 — three minutes after the bug report. #595 ships the same circular-variance formula plus 4 dedicated tests (including a regression test that explicitly asserts the old linear formula's pi² output on the wraparound input, so any future regression will fail the suite). #595 was merged in 60d1f… just now.

Thanks for the contribution. For next time, please check existing open PRs before opening a fix and open feature branches from your fork rather than main — that keeps your fork's main pristine for upstream sync.