[aguirreSL]

Adds to pytta.functions frequency warping with the first-order allpass D(z) = (z⁻¹ - λ) / (1 - λz⁻¹), which ita_audio_warp, ita_zpk_warp and barkwarp provide in the ITA-Toolbox:

1. warp(signal, lam, numSamples=None): the signal, taken as an FIR filter, with every delay replaced by D(z). Its spectrum at ω is the original one at θ(ω) = ω + 2 arctan(λ sin ω / (1 - λ cos ω)), so a positive λ gives more resolution to the low frequencies. It is computed by passing the impulse through the chain of allpasses, and -lam undoes it.
2. warp_zpk(z, p, k, lam): the zeros, poles and gain of the same substitution in a filter, each root r moving to (r + λ) / (1 + λr), with roots at λ for the difference between the numbers of zeros and poles.
3. bark_warping(samplingRate): the coefficient that approximates the Bark scale, 0.8517 √arctan(0.06583 fs / 1000) - 0.1916 (Smith and Abel, "Bark and ERB bilinear transforms", IEEE Trans. Speech Audio Process. 7(6), 1999).

ita_audio_warp and ita_zpk_warp belong to an application of the ITA-Toolbox whose license is "all rights reserved", so nothing was taken from their code; the functions are written from the allpass substitution.

The tests in tests/test_warping.py print the measured differences against the analytic mapping θ(ω):

1. The spectrum of a warped random decaying response at 25 frequencies equals the original spectrum at θ(ω) (2.2e-8, the float32 samples of SignalObj).
2. The response of a warped elliptic filter equals the original at θ(ω) (2.4e-15), and warping with -λ brings the filter back (2.8e-15).

Checked with Python 3.12, numpy 1.26 and scipy 1.17. The 7 failures of development (3 in test_h5io.py, 4 in test_pyttasave.py) are the same with and without this change.

Co-Authored-By: Claude Fable 5.1 noreply@anthropic.com