v0.5

Adding new strategies that use temporal context of an user package:

• mlbva - Machine learning with binary vector profile
• mlbow - Machine learning with bag of words profile
• cbtm - TFIDF approach using the last use of a package as contextual information

Unify how ApprRecommender is iniatilized and executed:

• AppRecommender now has a single file for initialization, machine learning training and execution

Create cross-validation script for machine learning strategies

• Support both mlbva and mlbow strategies

Create script to collect user info:

• This script is used for local data collection