More robust pandoc conversion

[marph91]

Affects all formats that use pandoc in the background. Tables are only converted to Markdown pipe tables if they are simple tables. Also lists aren't converted always.

Examples:

• In Zoho, some preprocessing is applied in HTML in order to create simple tables and lists.
• In Google docs (docx), the table is not converted.

Pandoc allows to use filters in lua that are fast and manipulate the AST directly. This would eliminate the need of another preprocessing step. The filter should:

• remove all divs (and other elements) from cells
• convert lists to - [ ] ... + linebreak
• convert newlines to some intermediate char (sequencs) and replace it in postprocessing