Corpus processing, statistics, and title detection improvement

This branch was originally to improve corpus processing, with the following changes :

• Adding a Corpus class and methods to retrieve useful stats.
• Integrating that class in corpus_processing.py
• Changing document.py and collection.py to retrieve info for statistics (mainly metadata)

This branch is up-to-date with synchro_eric, which also has numerous changes , such as :

• better logging mechanisms
• better title detection
• better table detection
• better text retrieval for logical sections
• etc...