Floati now reads HTML and CSS selectors exactly like the reference browser

Two of the most foundational parts of a browser engine - turning HTML text into a page, and working out which style rules apply to which elements - are now complete and match the reference browser exactly. On the HTML side the engine passes the entire reference conformance suite, every last case, including the newest form-control parsing rules and a notorious torture test of deliberately broken, badly-nested tags. On the CSS side, a fresh test that checks fifty-six kinds of selector against the reference browser's own answer now agrees with the reference browser on every one - which quietly fixed real rendering along the way: a whole family of selectors (things like "is this element empty", "is this control disabled", "does this contain that") used to match everything, so pages that leaned on them were mis-styled. On a large encyclopedia the search button no longer looks disabled, and reference numbers render cleanly; on other sites, rules that select an element based on what it contains now apply the way they should. None of this is flashy on its own, but it is the bedrock the rest of faithful rendering is built on, and it is now solid.