Measurement accountability for Gazepoint physiology

This tranche adds four conservative diagnostics.

compare_hrv_prv_devices() reports ICC(A,1), Lin CCC and Bland–Altman agreement per derived metric, carrying ECG/PPG and sensor-site provenance rather than producing a global interchangeability flag.

scr_responsivity_sensitivity() preserves the conventional SCR non-responder flag while estimating a Beta-Binomial response probability and retaining low-reactive participants for model-based sensitivity analysis.

physiology_validation_ladder() separates acquisition QC, analytical QC, construct evidence, within-person evidence and held-out-person generalization. Population/generalization claims require the final stage to pass.

ppg_topology_features() is opt-in and experimental. It summarizes H0 persistence lifetimes on a delay embedding as a structural waveform descriptor, never as a direct physiological surrogate, and carries a participant-grouped validation guardrail.

Methodological motivation: DOI 10.1186/s12872-026-06556-4, DOI 10.1016/j.biopsycho.2026.109336, DOI 10.3389/fnrgo.2026.1911259, and DOI 10.1016/j.measurement.2026.122168.