Updating feature extractor for handling special arg idx is complicated

[hagenw]

As shown in the docs, we can handle models that have different heads for different sexes, e.g.

def f0(signal, sampling_rate, idx, gender, f0_range):
    # extract mean f0 using a gender adapted range
    y = librosa.yin(
        signal,
        fmin=f0_range[gender[idx]][0],
        fmax=f0_range[gender[idx]][1],
        sr=sampling_rate,
    ).mean()
    return y, gender[idx]

interface = audinterface.Feature(
    ['f0', 'gender'],
    process_func=f0,
    process_func_args={
        'gender': gender,
        'f0_range': f0_range,
    },
)

The problem is that the gender argument is not really static but depends on the input data.
To update it you would need to create a new Feature object for each new database, which is not what we want for ONNX models, or you could update the process_func_args, which is ok, but they are hidden under interface.process.process_func_args.

So maybe we update the documentation by showing an example how to update them?