refactor(audio): Split audio module into modular files

[m96-chan]

Problem

Audio module is 5859 lines total across 3 files:

• native/ops/audio/audio_kernels.cuh (2023 lines)
• native/ops/audio/audio.cu (2009 lines)
• src/pygpukit/ops/audio.py (1827 lines)

Proposed Structure

Native (C++/CUDA)

native/ops/audio/
├── stft/
│   ├── stft.cu
│   └── istft.cu
├── mel/
│   ├── mel_filterbank.cu
│   └── mel_spectrogram.cu
├── features/
│   ├── mfcc.cu
│   ├── chroma.cu
│   └── spectral.cu
├── effects/
│   ├── pitch_shift.cu
│   ├── time_stretch.cu
│   └── filters.cu
├── vad/
│   └── vad.cu
└── common/
    └── audio_common.cuh

Python

src/pygpukit/ops/audio/
├── __init__.py
├── stft.py
├── mel.py
├── features.py
├── effects.py
└── vad.py

Benefits

• Each file < 500 lines
• Clear separation by audio processing stage
• Easier maintenance and testing

Related

• refactor(native): Split ops_bindings.cpp into modular files #131 (binding split)
• refactor(native): Split nn.cu into modular files (aligned with #131) #133 (nn.cu split)

[m96-chan]

Completed in PR #168