Evaluate alternatives to storing executed notebooks in git

Doing widget state cleanup today:

$ git diff --stat
 docs/source/examples/Analyzing the Parker Solar Probe flybys.ipynb                  |  4295 ------------------------
 docs/source/examples/Catch that asteroid!.ipynb                                     |  2655 ---------------
 docs/source/examples/Exploring the New Horizons launch.ipynb                        |  1446 --------
 docs/source/examples/Going to Jupiter with Python using Jupyter and poliastro.ipynb | 11647 ----------------------------------------------------------------
 docs/source/examples/Using NEOS package.ipynb                                       |  6689 +------------------------------------
 docs/source/examples/Visualizing the SpaceX Tesla Roadster trip to Mars.ipynb       |   141 +-
 6 files changed, 106 insertions(+), 26767 deletions(-)

This was just the last straw, but storing executed notebooks takes a lot of space and makes code review more difficult. We should evaluate whether things like jupytext are better suited for this.

[jorgepiloto]

When implementing #828 and after #976 it is clear that some notebooks do not seem to be updated even if a make clean && make html is executed. In ddition to that, reviewing modifications is difficult when it comes to *.ipynb as @astrojuanlu claimed.

I did a small ship locally from *.ipynb to *.py using Jupytext and found these advantages apart from previously ones:

• Notebooks are forced to run always, no matter if they do not raise API changes related errors.
• The recently implemented nbsphinx-gallery works without problem together with 3D plotting utilities.
• Syntax is easy and it is possible to re-convert to notebooks whenever it is desired.

[astrojuanlu]

@GorgiAstro if you notice weird things with the TOC extension, please let us know!