[php-src] Issue #7863: Replace Libxml2 by Gumbo or Lexbor for HTML parsing

Issue: https://github.com/php/php-src/issues/7863 Author: woho ### Description DOMDocument->loadHTML uses Libxml2 and is fundamentally broken. Valid HTML documents exist that cannot be parsed correctly and there are no simple workaround that work in all cases. Specifically: - Closing HTML tag inside script contents, which is not a rare occurrence, break the parser. Fixing this for all possible cases by pre-processing is not possible without complete tokenization of the document, because strings like "<script>" can occur in attribute values. - Encodings are poorly handled. Libxml2 only reliably supports UTF-8 and ISO-8859-1, but other encodings such as ISO-8859-2 are popular for some languages. Workarounds are very dirty, the easiest one is to identify the encoding by parsing the document twice and convert the whole document using mb_convert_encoding to 'HTML-ENTITIES'. There are two good HTML parsing libraries written in C, [Gumbo](https://github.com/google/gumbo-parser) and [Lexbor](https://github.com/lexbor/lexbor). Their performance is similar. [For Gumbo, a PHP extension already exists](https://github.com/layershifter/gumbo-php), which integrates seamlessly with DOMXPath. It would be great to have a properly working native solution for loadHTML in PHP, after all it is a web development language.