[aguirreSL]

Adds pytta.din45681 with the tonal audibility of DIN 45681:2002, as ita_tonality in the ITA-Toolbox (applications/Psychoacoustics).

tonal_audibility(signal) takes one channel of sound pressure in Pa and returns the largest audibility DeltaL in dB, the frequency of that tone and a list with each tone found: its frequency, critical bandwidth, LS, LT, LG, av, DeltaL and DeltaL of its critical band. The method follows ita_tonality step by step:

1. The A-weighted power spectrum is averaged over Hann windowed blocks with lines of 2 Hz.
2. For each line from 100 Hz to 8192 Hz, the masking level LS of its critical band, dfc = 25 + 75 (1 + 1.4 (f/1 kHz)^2)^0.69 Hz, is the energy mean of the lines not more than 6 dB above LS, iterated until it settles.
3. A region of lines 6 dB or more above LS is a tone candidate at its highest line. Its level LT adds the neighbour lines within 10 dB of the peak.
4. DeltaL = LT - LG - av, with LG = LS + 10 lg(dfc/df) and av = -2 - lg(1 + (f/502)^2.5). A candidate wider than 18 (1 + 0.001 f) Hz, or with a flank less steep than 24, is no tone.
5. Tones in one critical band are summed.

The code is a port, so the licence of the ITA-Toolbox (BSD, four clauses) is reproduced in the module header, with the acknowledgment its third clause asks for. Three points differ from the MATLAB code, and the docstring states them:

1. ita_tonality labels FFT line k with the frequency (k + 1) df, so every frequency it reports is 2 Hz high. This module uses k df.
2. Its levels are 6 dB low, for tones and noise alike, so DeltaL is the same. This module gives absolute levels.
3. Its search for a tone region uses LS = 0 below 100 Hz and would read past the end of LS above 8192 Hz. This module stops the search at both ends.

The tone adjustment K_T of the standard is not computed, as in ita_tonality.

The tests in tests/test_din45681.py print the measured differences:

1. A 1 kHz tone in white noise gives the DeltaL of the formula within 0.19 dB. The masking level of a noise estimated from 19 blocks scatters by up to 0.34 dB over noise seeds, so the tolerance is 0.5 dB.
2. Parity with ita_tonality on golden data from tests/golden/make_din45681_ita.m (MATLAB R2026b, same samples): tones at 400 Hz, 1 kHz and 3 kHz in noise and two tones in one critical band give the DeltaL of ITA within 0.005 dB, the same number of tones, and frequencies 2 Hz below those of ITA.
3. Noise alone has no tone.
4. Invalid input is rejected.

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