122Implement a simplified HTML parser and DOM tree builder

Without DOMParser or innerHTML, implement parseHTML(html), which returns a tree:

• the root is { type: "root", children };
• elements are { type: "element", tag, attributes, children } (tag names lower-cased, attributes a plain object);
• text is { type: "text", value } (with entities decoded).

Also implement toHTML(node), which serialises a tree back to HTML (escaping text and attribute values).

Support:

• nested elements and text;
• attributes with double quotes, single quotes, no quotes, and boolean attributes (<input disabled> gives disabled: "");
• void elements (area, base, br, col, embed, hr, img, input, link, meta, source, track, wbr) and self-closing syntax (<x />);
• comments (<!-- … -->) and <!doctype> are skipped;
• entities & < > " ' and numeric &#NNN; / &#xHH;;
• <script> and <style> contents are raw text (no tags parsed inside);
• error recovery: a closing tag closes the nearest matching open element (implicitly closing anything inside it); stray closing tags are ignored; unclosed elements are closed at the end.