Formalize unicode/string handling schema

[hxu]

lxml has some issues with unicode strings and encoding declarations. At the moment, the code kind of converts between string and unicode in an ad-hoc way. We should formalize a framework for handling this (e.g. strings should be converted to unicode as early as possible).

See http://stackoverflow.com/questions/15302125/html-encoding-and-lxml-parsing and http://lxml.de/parsing.html#python-unicode-strings for discussion

[hxu]

Also this presentation: http://farmdev.com/talks/unicode/