score each section

[bmoscon]

instead of returning true / false for each classified section, record a score based on each heuristic. for example:

anchor density too high? score - 1, else score + 1. items with positive scores are increasingly likely to be content, and with negative scores are increasingly likely to be boilerplate.