[feature] AsymKNN for user and item based KNN

[mkhe93]

• Implemented user-based and item-based asymmetric KNN versions as described in Aiolli (2013)

• Updated the documentation and introduced the following parameters:

  • knn_method (str): [item, user]
    Specifies whether to use user-based or item-based KNN
  • k (int):
    Neighborhood size
  • alpha (float): [0, 1]
    Weight parameter for asymmetric cosine similarity
  • beta (float): [0, 1]
    Final score balance parameter
  • q (int):
    Exponent for the "locality of scoring function"

Created this algorothms for my researches. I'm happy for feedback and improvements!