Investigate automatically comparing mwph to parsoid output

[legoktm]

We should investigate building a tool or test suite of some kind that compares the parsed mwph tree to Parsoid's DOM output. We would map nodes to the Parsoid dom spec

I imagine something like this with a large dump like en.wp could have caught #120.

[jayvdb]

The two trees would be very different for almost every page where the differences are likely to be interesting. e.g. we would need to remove lists from both trees, to work around #46.