pin how numeric character references in the c1 range extract issue 76 reported that entities like the ellipsis came out as the raw code point instead of what a browser shows. that is fixed now in libxml2 but nothing in the suite holds it there. these cover the whole range against the table python implements so a regression is caught on 27 of the 32 code points. the xml one is pinned too because the literal code point is correct there and should not be changed to match html.