When inline Markdown elements are nested, each nested level contains its
own placeholder. Therefore, when unescaping, the regex substitution needs
to be run against each nested level. As this is nested inside the match, it ensures
no more than one non-matching regex will be run in each instance.

Closes #1571.

Fixes #1573.

The `updatepos()` override was resetting the parser position to `0`,
causing infinite loops when multiple bogus/unclosed comments
appeared in the input (e.g., `</` or `<!--` followed by non-standard
characters). Now tracks the position of the triggering `<` and
advances past it instead of rewinding to the start of the buffer.

Fixes #1578.

* update changelog

Fixes #495.

This reverts commit 07dfa4e.

- Have `parse_comment` handle bad `<!--` cases.
- `handle_comment` only needs to handle the `</` case for comments.
  Provide a more reliable way to calculate the index to check for this
  case.
- Ensure `override_comment_update` is reset, just in case.

Resolves #1586

* Fix spelling

* Add changelog entry