feat: EmbodiedConductor — dual-mode bio-musical orchestrator

[Copilot]

Summary

Implements the complete Embodied Conductor system for BrainJam — a dual-mode bio-musical orchestrator that bridges human physiology (EEG, PPG/HRV, gestures) with real-time music control.

New Modules

File	Description
performance_system/embodied_conductor.py	Top-level orchestrator with CREATOR/MAESTRO dual mode, 16-bar latent phrase memory, MIDI rubato engine, and unified ConductorState event
performance_system/biofeedback/ppg_hrv_pipeline.py	Butterworth band-pass filter (0.7–3.5 Hz) → peak detection → rolling RMSSD → rubato control output
performance_system/mapping_models/dissonance_index.py	EEG beta/alpha ratio mapped via sigmoid to harmonic tension (Creator) or velocity/articulation pressure (Maestro)
performance_system/embodied/predictive_gesture.py	OLS-based 500 ms latency-compensation predictor for MediaPipe hand positions; adds hand-closure scalar
performance_system/signals/realtime/mobile_sensor_bridge.py	Async WebSocket server (port 9001) ingesting phone-camera PPG frames with timestamp normalisation and jitter filtering
docs/architecture/embodied_conductor.md	Architecture reference with full mathematical annotations for PhD proposal documentation

Changes to Existing Files

• performance_system/realtime/event_bus.py — Added CONDUCTOR_STATE, PPG_UPDATE, MODE_CHANGED event types
• performance_system/biofeedback/__init__.py — Export PPGHRVPipeline
• performance_system/embodied/__init__.py — Export PredictiveGestureController, PredictedGestureState
• performance_system/signals/realtime/__init__.py — Export MobileSensorBridge

Tests

• tests/test_embodied_conductor.py — 51 tests, all passing
  • PPGHRVPipeline: RMSSD computation, Butterworth filter, simulate mode, quality gating
  • DissonanceIndex: band ratio normalisation, EMA smoothing, articulation thresholds
  • PredictiveGestureController: OLS prediction, hand closure, z_max boundary
  • LatentPhraseMemory: store/recall/mutation, cosine similarity threshold
  • EmbodiedConductor: dual-mode ticks, mode switching events, async lifecycle, phrase recall
  • Regressions: existing GestureController, HRVProcessor, and EventBus unchanged

Mathematical Reference (Quick Summary)

PPG → RMSSD:   rubato_factor = clip((RMSSD_ms − 15) / (80 − 15), 0, 1)
EEG tension:   D = sigmoid(k · (β/α − μ)),  k=2.5, μ=1.2
Gesture pred:  p̂(t+h) = v̄·(t+h) + p₀  via OLS (h=500 ms)
Rubato engine: BPM_mod = BPM_base / (1 + 0.25·(2R−1))
Phrase recall: cos_sim(v_current, v_stored) ≥ 0.92