Make classic microformats support much stricter

[barnabywalters]

If any mf2 classnames are found in a document, don’t convert any classic mf classnames

E.G. currently parsing this URL with classic support turned on produces bad results due to classic mf classnames being on different elements to mf2 classnames — with stricter checking this would not be a problem.

[aaronpk]

Here is another example of overly aggressive parsing:

http://pdx.esri.com/blog/2013/10/17/introducting-mapattack/

The "description" values should not be in the parsed result.

[aaronpk]

Because of the overly aggressive parsing, I propose removing the default of "parse legacy microformats" for now.

[barnabywalters]

Resolved in v0.2.0 with improved (more spec-compliant) classic mf support. I’m leaving convert classic on by default to help root out bad markup, it shouldn’t be so much of a problem now.