[feature] ALS Matrix Factorization using External Library (implicit)

[mkhe93]

NOTE: this feature uses an external library implicit, similar to the implementations of LightGBM and XGBoost.

Implemented external ALS matrix factorization algorithm using ALS implicit. The original work has been done by Koren in Collaborative Filtering for Implicit Feedback Datasets and performance optimization in Applications of the Conjugate Gradient Method for Implicit Feedback Collaborative Filtering.

Updated the documentation and introduced the following parameters:

• regularization (float): 0.01
  The number of latent factors to compute

• embedding_size (int): 64
  The regularization factor to use

• alpha (float): 1.0
  The weight to give to positive examples.

The usage of implicit through RecBole brought several benefits to my research, which I would like to share with people who may also be interested. I welcome feedback and suggestions for improvement!