Refactor sound generation functions to remove user_id parameter and adjust GPU duration

Implement GPU quota management and user registration for sound generation

Add support for GPU acceleration in sound generation and update dependencies

Update output file naming in sound_generator.py to include timestamps for unique identification

Remove unnecessary device transfer for sound and music models in sound_generator.py

Fix model loading in sound_generator.py to access the model attribute directly

Add device information print statement in sound_generator.py

Refactor sound_generator.py to load models on GPU if available and streamline audio generation process

Add initial project structure with FastAPI and audio generation functionality