Improved usabilty

supported features:

• computation of co-occurrence matrix
• in-library encoding using pre-trained models
• in-library compression/expansion of the oneshot vector

known issues:

• length of the covariate vector can only be equivalent to the length of the ones in df_M
• does not support a sparse vector data structure, necessary for highly scalable scenarios