The `get_extremum_channels` should have an `auto` option.

Currently the get_extremum_channels (and many other functions) require a peak_sign. However, since the input of all these functions is usually a WaveformExtractor, we could add the option to automatically estimate the peak_sign from the templates.

e.g.:

if peak_sign is None:
    mean_template_value = np.mean(all_templates)
    peak_sign = "neg" if mean_template_value < 0 else "pos"
# or we could use the skewness of the all_templates distribution, which should have a negative (or positive) tail

What do you guys think?

Locate get_extremum_channels and the other functions that require peak_sign, then inspect how they receive a WaveformExtractor and access templates. The proposed behavior is not settled between using the mean template value or distribution skewness, so confirm the intended estimation rule and define tests showing that the auto option selects the expected peak sign.