Findings:

- We can download the following formats:

    html
    txt
    rtf
    odt
    pdf
    docx
    epub

- Pandoc only supports reading from

    html
    odt
    docx
    epub

- Inspecting and comparing the native conversion, here are some observations:

  * Both the HTML and epub outputs similarly produce crappy representation that
    incorporates a lot of stylistic elements. This leaves odt and docx.

  * Surprisingly, odt parsing does not yield a whole lot of useful structure.
    For instance, the section headers aren't even parsed as such, but rather
    contain anchor markers. Not great.

  * So docx is the most useful here.

    * It does not produce a title. We may have to download multiple formats in
      order to extract different parts from different formats.

    * It has extraneous blockquotes.

    * The code blocks aren't detected as such and whitespace in front of account
      names is lost.

Converting out to HTML, we see

- That the output needs to be spliced in an HTML wrapper with a UTF-8 encoding.
