[aguirreSL]

Adds pytta.smooth_frequency(signalObj, nthOct=3): fractional octave smoothing of the magnitude spectrum, phase kept, returning a new SignalObj.

The magnitude of freqSignal is interpolated linearly onto a logarithmic grid of 1 cent spanning the 20 octaves below Nyquist, box averaged over 1200 / nthOct cents (made odd, shrinking symmetrically at the ends of the grid) and interpolated back onto the bins with a not-a-knot cubic spline. The DC bin is copied unchanged. This is the LogFreqOctave1 smoothing of ita_smooth in the ITA-Toolbox (RWTH Aachen), written from its documented behaviour; no ITA code is copied.

The first commit fixes a bug that this feature needs: building a SignalObj from a spectrum (domain='freq') failed for signalType='power' because _ifft multiplied the whole half spectrum into the AC slice (shape mismatch). The fix indexes [1:, :] on both sides; the round trip time -> freq -> time now closes at the float32 floor for both signal types.

Tests (tests/test_smooth_frequency.py, 6 tests):

• a flat magnitude with random phase comes back unchanged (magnitude only, phase kept)
• a single bin spreads into a plateau of the predicted height area / width over the predicted band, flat to float32 and zero beyond it, at 1, 1/3 and 1/12 octave
• DC and Nyquist bins are kept
• new object, metadata kept, channels smoothed independently
• rejection of invalid nthOct and input type
• sample by sample comparison with the ITA-Toolbox at 1, 1/3 and 1/12 octave. Reference data: tests/golden/smooth_frequency_ita.mat, generated by tests/golden/make_smooth_frequency_ita.m with MATLAB and the ITA-Toolbox. Measured difference: 2.7e-7 of the spectrum peak, the float32 floor of SignalObj.

Note for reviewers: PyTTa and the ITA-Toolbox scale the DC and Nyquist bins differently from the AC bins. The reference signal has those two bins at zero so that the comparison pins the smoothing itself; the function smooths freqSignal as stored.

Checked with Python 3.12, numpy 1.26, scipy 1.17. The existing tests behave as on development: the 7 failures there (3 in test_h5io.py, 4 in test_pyttasave.py) are the same with and without this change. Independent of PR #49.

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