introduced math.nan

I introduced here math.nan to be indenpendent from pandas in case formatted parsing is not required. I think it's still more conveninent to have nan instead of 0. Zero can also be a valid filesize while nan indicates invalid filesizes, e.g., for directories. It's also more convenient when workind with pandas... (dropna, isnull, etc...). I also added some more tests and an example notebook.. Let me know if you are ok with this...

This merge request comes with:

• math.nan
• added test_utils.py
• added examples.ipynb notebook.