1. _gettoks(): lex styles like scripts (no entities); fix ts leak.
2. getscriptdata(): add findtag argument to accomidate styles and fix comment lexing.
3. amperstand(): change entity recognition to parse unicode character references
and determine the end of reference as per rfc.

changes 1-3 will allow abaco to correctly display wikipedia and yahoo.co.jp.
it also improves the output of htmlfmt.