[rb-general] Addresses in (I)Python output

[Rebecca N. Palmer]

I agree that having the search-and-replace code in a central place 
(similar to the existing fix for memory addresses in API documentation 
[0]) rather than per-package is a good idea.  However, that place should 
ideally be IPython/Jupyter not Sphinx, as this also affects standalone 
.ipynb files or files processed with other converters.

As an estimate of the scale of this problem (I haven't checked how many 
of these actually have addresses in them),

Debian packages containing .. ipython:: in the source: python-geopandas, 
statsmodels, python-h11, pandas, python-pbcommand, matplotlib, 
matplotlib2, colorspacious, python-cycler, yade, python-xarray, patsy, 
ipython, pikepdf

Debian packages with .ipynb files in the binary: deal.ii, matplotlib, 
matplotlib2, pymvpa2, statsmodels
but there are more (e.g. theano, though I don't think that one has this 
issue) that have .ipynb files in the source and use jupyter-nbconvert / 
pandoc / etc (not Sphinx) to convert them to HTML in the binary.

[0] 
https://github.com/sphinx-doc/sphinx/pull/1694/commits/d24bd73d0ce8071459a1056691a1934e1ca12194