Improve assessment for the dissimilar glyph pairs

[MariannaPaszkowska]

Hey there! Halfkern is currently providing some overkern suggestions for the "-J" and "C-". Maybe there is a way to improve this.

[behdad]

Yeah I realized that with a Serif font. I think I like the "C-" one...

Note that of you pass -l to the kern_language.py it will only process letters, no punctuation.

[behdad]

overkern

I need to change algorithm slightly to handle those better.