feat(sensing-server): adaptive person count — RollingP95 + dedup_factor runtime API

[schwarztim]

Motivation

Person counting in v2/ uses fixed-scale feature normalization which works well in calibrated environments but degrades when signal characteristics drift across rooms, interference levels, or hardware. Two improvements here are deployment-neutral:

1. RollingP95 adaptive normalizer

compute_person_score() previously normalized features with hard-coded denominators (variance/300, motion_band_power/250, spectral_power/500). When live ESP32 values exceed those limits the normalized inputs clamp to 1.0 and dynamic range collapses. RollingP95 is a streaming P95 estimator (600-sample / ~30 s sliding window) that self-calibrates to whatever feature distribution the deployment produces. Cold-start (< 60 samples) falls back to the legacy denominators so day-0 behaviour is fully preserved.

2. dedup_factor runtime API

Exposes the multi-node cluster deduplication divisor via REST so deployments can tune to their environment without rebuilding. Includes an auto-tune endpoint that derives the optimal dedup_factor from a known person count (calibration mode). Config persists across restarts in data/config.json.

Explicitly NOT included

This fork also has additional ISTA lambda tuning (specifically lambda=5.0) for its local 8×8×4 babycube grid. Those values are deployment-specific and intentionally not included in this PR — they would degrade person-count quality on different room geometries. This PR keeps upstream's existing lambda: 0.1 default.

Changes

• v2/crates/wifi-densepose-sensing-server/src/main.rs

  • New RollingP95 struct + impl (ADR-044 §5.2)
  • New RuntimeConfig struct + load_runtime_config / save_runtime_config (ADR-044 §5.3)
  • AppStateInner: added p95_variance, p95_motion_band_power, p95_spectral_power, dedup_factor, data_dir fields
  • compute_person_score() signature: &FeatureInfo → &AppStateInner + &FeatureInfo (adaptive denominators)
  • All 3 call sites updated; P95 push calls added before each scoring call
  • New REST handlers: config_get_dedup_factor, config_set_dedup_factor, config_set_ground_truth
  • Routes registered: GET/POST /api/v1/config/dedup-factor, POST /api/v1/config/ground-truth
  • 5 unit tests in rolling_p95_tests module

• v2/crates/wifi-densepose-sensing-server/src/multistatic_bridge.rs

  • fuse_or_fallback() gains dedup_factor: f64 parameter; fallback switches from max() to ceil(sum / dedup_factor)
  • Test call site updated

Test results

• cargo test --workspace --no-default-features: 1636 passed, 0 failed (includes 5 new RollingP95 unit tests)
• python archive/v1/data/proof/verify.py: VERDICT: FAIL — pre-existing on origin/main (numpy/scipy version drift); not caused by this PR

Notes

• The auto-tune endpoint accepts a known person count and adjusts dedup_factor to match observed cluster count. Useful for calibration during install.
• RollingP95 is a generic primitive — could be reused for other adaptive thresholds in future.
• lambda: 0.1 in tomography.rs is untouched.