Improvements of PBstat and creation of regression tests

[HubLot]

Hi,

In this PR, you will find a first step in the modularization of PBstat. I changed a little bit the PBstat.py:

• remove the old function needed for R graphics (array_to_string) with it's call. It wasn't used anymore
• Make the subprocess lines for weblogo compatible with Python 3:

Since in Python, there is a clear distinction between bytes and str, the line :

 (out, err) = proc.communicate(transfac_content)

was not working. I change it to specifiy the coding:

 (out, err) = proc.communicate(transfac_content.encode('UTF-8'))

I also created a few regressions tests :

• check if the image files (png and pdf) are created and are not empty.
• for --neqoption, assess the Neq file is the same as the reference. Hence, few references files have been added to test_data.
• update the travis file for matplotlib and weblogo dependencies.