ENH: Refactor writer/translators into specific targets ipynb, html, pdf

[mmcky]

We should refactor the current translate_code and translate_all into specific end target writers to improve this code base. It is currently confusing to have all the states included in a single writer depending on a wide set of options.

I propose:

1. translate_code as the base object used to write notebooks that just contain code blocks
2. translate_ipynb for as the primary for producing ipynb files that has an emphasis on readability and supports markdown as much as possible
3. translate_html for writing ipynb files used in the production of the website and favours using html objects in the notebooks
4. translate_pdf for writing ipynb files used in the production of pdf files.

[mmcky]

More detailed discussion of this is

https://github.com/QuantEcon/sphinxcontrib-jupyter/wiki/Extension-Refactor-and-Redesign-(November-2019)