Release 3.6

Changed

Refactor TOC Sanitation

• All postprocessors are now run on heading content.
• Footnote references are now stripped from heading content. Fixes #660.
• A more robust striptags is provided to convert headings to plain text.
  Unlike, the markupsafe implementation, HTML entities are not unescaped.
• The plain text name, rich html, and unescaped raw data-toc-label are
  saved to toc_tokens, allowing users to access the full rich text content of
  the headings directly from toc_tokens.
• The value of data-toc-label is sanitized separate from heading content
  before being written to name. This fixes a bug which allowed markup through
  in certain circumstances. To access the raw unsanitized data, retrieve the
  value from token['data-toc-label'] directly.
• An html.unescape call is made just prior to calling slugify so that
  slugify only operates on Unicode characters. Note that html.unescape is
  not run on name, html, or data-toc-label.
• The functions get_name and stashedHTML2text defined in the toc extension
  are both deprecated. Instead, third party extensions should use some
  combination of the new functions run_postprocessors, render_inner_html and
  striptags.

Fixed

• Include scripts/*.py in the generated source tarballs (#1430).
• Ensure lines after heading in loose list are properly detabbed (#1443).
• Give smarty tree processor higher priority than toc (#1440).
• Permit carets (^) and square brackets (]) but explicitly exclude
  backslashes (\) from abbreviations (#1444).
• In attribute lists (attr_list, fenced_code), quoted attribute values are
  now allowed to contain curly braces (}) (#1414).