Standardization of API provided to the Python runtime

With the new Python kernels which are now arising, it may be interesting to define a scoped API that a Python kernel must make available for the Python runtime.

For example, `xeus-python` monkey-patches `IPython.display`, `IPython.clear_output`, etc... I just opened https://github.com/jupyter-widgets/ipywidgets/issues/2665 in ipywidgets to list the IPython and ipykernel APIs that we are using and may be included in this limited API.

Instead of monkey-patching and using the IPython module, a way to go may be for the kernel to dynamically create a "kernel" python module that has all these functions...

This somewhat goes along the lines of @Carreau's remark on points of convergence between ipykernel and xeus-python:

> If users needs to switch between ipykernel and Xeus-python depending on wether they want to use a given magic or the debugger the experience will be sub-par. If we have to choose, a clear distinction of features would be better than two almost similar projects.

Start by reviewing the linked ipywidgets issue 2665 and comparing the IPython and ipykernel APIs used by xeus-python. Define the scope and compatibility requirements for a shared Python-kernel API, including whether a dynamically created kernel module is appropriate; the work is done when the API boundary and adoption path are agreed.

Scritto dal modello di indicizzazione a partire dal testo della issue.