Further improve the double percent format

[mwouts]

The double percent format seems to have a very nice potential as a standard text representation for Jupyter notebooks. It is already understood by many editors, some of which support other languages than Python.

Below is a list of desirable improvements for jupytext's implementation of the format:

• Read cells from scripts generated by jupyter nbconvert. Currently most editors that can read the double percent scripts already do that, i.e. cells may start with In [21], etc.
• Remove up to two blank lines at the end of cell (these blank lines may appear before functions or classes, when Python code is reformatted).
• Improve support for Spyder sub cells: cell title goes to Jupyter cell metadata (with no initial % in it), and cell depth is stored in another metadata.
• Allow different comment chars for different languages (; or ;; for scheme or lisp, // for c++, etc).
• Implement an option for commenting Jupyter magics in the script. Depending on the interpreter, the user may want or not to execute these commands.

[mwouts]

Available in v0.8.0