Class WaveletFeatures is confusing

[Catarina-Alves]

There are several things in the code that are messy: add-ons to add-ons, missing docstrings, non-clear names, missing functionalities (or functionalities that are there but that are hard to find) and the code does not follow the convention (which is natural as it was written before it).

I propose to improve this class, solving most of these problems (including issues #72, #75, #76, #77, #78 and #112).
This is important for me now as I am verifying the features and it is harder than it should be.

I am going to change mostly what I am using to verify the features but if you have requests/ comments regarding specific parts, please write them here.

[Catarina-Alves]

This should resolve issues #177, #152, #148, #112, #78, #77, #76, #75, #85, #70
It might also solve issues #107, #91 (in parte), #65

snfeatures restructuring - general:

• Remove legacy code - any code that is commented out is erased
• Add more documentation to some functions (e.g. fit_sn)
• Have more descriptive names (e.g. currently wavelet_decomp performs wavelet decomposition on a single light curve and extract_wavelets performs in all dataset but the names don’t show the similarities/ differences)
• Be more consistent in naming and overall have better names for variables and functions

snfeatures restructuring - desired functionalities (some already exist):

• from GP -> calculate full wavelet decomposition
• from GP -> calculate reduced wavelets
• from full wavelet decomposition -> calculate reduced wavelets
• from full wavelet decomposition -> reconstruct light curves
• from saved PCA -> compute reduced wavelet decomposition with different number of components
• from saved PCA -> project new data into specific number of components
• from reduced wavelets -> reconstruct full wavelet decomposition (lossy)
• from reduced wavelets -> reconstruct light curves
• from reduced wavelets -> compute reconstruction error

Other:

• Remove __future__ import
• Only keep the SVD eigendecomposition because it is the most efficient method

[Catarina-Alves]

Solved with #219.