feat: PyGaborSTM v1.0 - STM implementation with validation

[JaredChilders]

Summary

Complete implementation of the spectro-temporal modulation (STM) framework based on Chi et al. (1999, 2005) and Bellur & Elhilali (2017).

What's Included

Core Implementation

• Auditory spectrogram pipeline (gammatone filterbank → transduction → lateral inhibition → integration → compression)
• 2D Gabor filterbank with configurable rates and scales
• Rate-Scale-Frequency (RSF) representation

Validation

• Validated against MATLAB NSL Toolbox
• Chi99 Matched Filter MTF
• Chi05 complex sounds (tone, harmonic, noise, speech, music)
• Moving ripple (mvripfft) selectivity tests

Documentation

• MkDocs + Material theme
• Auto-generated API docs from docstrings
• Getting started guide
• Example notebooks

Backend

• NumPy/CuPy backend abstraction
• GPU support via CuPy (optional)

Known Limitations

• High rates (32 Hz): σ_t ≈ 4 frames, reduced selectivity
• Extreme scales (0.25 cyc/oct): σ_f = 2 octaves, filter overlap
• Optimal range: 2-16 Hz rate, 0.5-4 cyc/oct scale

TODO

• Add config options for gammatone filters (filter order, ERB scale factor)
• Configurable compression function (log1p vs cbrt)

References

• Chi, T., Ru, P., & Shamma, S. A. (1999)
• Chi, T., et al. (2005)
• Bellur, A., & Elhilali, M. (2017)