on the ''raw'' text, which has not been HTML-escaped.
There's simply one more regular expression which does the
HTML escaping as part of the wiki processing.
This has two advantages:
* no need to unescape again, in several circumstances
* simple regular expressions can be built to handle the "&" character
The latter enables us to fix #2455.
On the other hand, care must be taken by the various _formatter methods
to correctly HTML escape the matched text groups.