1. Fix output of jupyter notebooks saved in VSCode

VSCode uses the newer NBFormat Version 4.1 [1], which allows
the output text entry to be a simple string.

The function `rawCode` assumed NBFormat Version 3.2 [2], which requires
the output.text object to be an array.

This commit adds a special handling if the type of output.text is a
string, then we don't need to use `join()` and can just return the
string.

Fixes: https://gitlab.com/gitlab-org/gitlab/-/issues/212837

[1] https://nbformat.readthedocs.io/en/latest/format_description.html
[2] https://ipython.org/ipython-doc/3/notebook/nbformat.html

2. Handle katex render error in ipython markdown

When a formula brings katex to its knees the whole cell is dropped.
Instead of just dropping continue on and leave the unrendered formula in
the cell.

Partially fixes: https://gitlab.com/gitlab-org/gitlab/-/issues/216744

3. Render listitems with Katex in ipython markdown

Add a renderer callback for list items. Not only paragraphs should have
the honor to have math formulas rendered with Katex. Also listitems of
bullet points and enumerations should have this honor.

Fixes: https://gitlab.com/gitlab-org/gitlab/-/issues/216741

4. Get the tick `'` back to pass it to Katex

When rendering IPython notebook markdown cells the formulas are passed
to Katex. Before that the raw string is passed to marked.js to handle
HTML forbidden characters. One of these characters is the `'`. This
character is often used in math formulas and the replacement `&#39` is
not well received by Katex.

Fixes: https://gitlab.com/gitlab-org/gitlab/-/issues/216744