There's a pretty good implementation of this over at the nltk repo, see :
readabilitytests.py
I'm pretty sure that this could be used to provide an interface for doing the calculation, however the tricky bit is more in ensuring that you only compute the score off of paragraph text out of the book chapters. I'm not too sure how this would be managed within the calibre plugin structure, however at the very least I'm sure you could get an interface to paste the text into, compute the score and store it.