The classname prefixes are always ASCII, so mb_ functions are not needed

It's unclear what problems the function was originallymeant to solve;
it is equally unclear the problems still exist

This is the result of extension testing against PHP 5.6 through
PHP 8.2 in Linux and Windows to determine how encoding detection
actually works in DOMDocument

As UTF-8 is now heuristically detected, no explicit encoding declaration
is required

Neither is likely to appear in a real HTML document; x-user-defined is
not actually supported properly, but it's unlikely to be used for HTML

All new lines are covered, and an effort was made to cover as many
permutations of PCRE matches as practical