Wave-PCS with manual and statistical FIR filter design methods

1. Design perceptual contrast stretching (PCS) filters for audio signals with two modes available: 'manual' uses the default gains from the official PCS repo; 'statistical' circumvents the log-1-p non-linearity with impulse response approximation.
2. Example PyTorch code to demonstrate how the designed FIR filter can be efficiently implemented in deep learning pipelines.
3. Code snippets to validate the wave-PCS results and the original spectral-PCS results.