Use MaxAbsScaler for some Multiplicative Signal Transformations

[antoinecarme]

PyAF uses a MinMaxScaler for all transformations by default. This introduces zero values in the scaled signal.

https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html#sklearn.preprocessing.MinMaxScaler

Multiplicative transformations (RelativeDifference, BoxCox) should work better by using a MaxAbsScaler whic his more adapted for positive signals.

https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MaxAbsScaler.html#sklearn.preprocessing.MaxAbsScaler

[antoinecarme]

This "feature" was introduced when twaeking PyAf 5.0-Fixes branck. A commit is already made :

https://github.com/antoinecarme/pyaf/commit/9420fc6c6c794a306258790ef6c97ce5f8166a4b

This issue was created to keep track of this significant change as a part of another issue :

234