Implement musical candidate ranking and feedback-driven learning

Refactor app.py into logical sub-modules

Implement explicit tempo policy and transparent volume headroom guards

Implement core DJ engine: transition optimization, cue learning, and set rendering

v4: no double-kicks (5kHz split+sneak), xcorr beat align, pre-drop mix-in (structure.py)

Iter 2: smooth A fade during drop, no outro exits, late-exponential curves (structure.py)

Beat-locked transitions, shorter durations, B enters at buildup not intro (structure.py)

Structural rewrite: energy-based segmentation, mix-out at breakdown, mix-in at intro, transition lands on drop (structure.py)