feat: add procedural environmental audio synthesis to Gen mode

Implements a complete audio system for 3D scene generation with:

- FunDSP-based procedural sound synthesis with lock-free Shared parameters
- Ambient soundscapes: wind, rain, forest, ocean, cave, stream
- Spatial audio emitters: water, fire, hum, wind, custom waveforms
- Auto-inference system that detects entity names and assigns sounds
- Distance-based attenuation and stereo panning
- cpal audio output with separate audio thread for graph management
- 4 new LLM tools: gen_set_ambience, gen_audio_emitter, gen_modify_audio, gen_audio_info

Uses 3-thread architecture: Bevy main thread → audio management thread (FunDSP Net frontend) → cpal callback (Net backend) for glitch-free audio without blocking main thread or GPU operations.

Author: yiwang

Cargo.lock

@@ -34,3 +34,7 @@ gltf-json = { version = "1", features = ["names"] }
 
 # Path resolution
 shellexpand = { workspace = true }
+
+# Procedural audio (environmental soundscapes)
+fundsp = "0.20"
+cpal = "0.15"