prob_classify for decision trees

[alexrudnick]

It would be extremely useful to be able to use DecisionTreeClassifier for
producing statistical decision trees, e.g. having prob_classify implemented.

It could be done by storing in each node the number of training examples
that passed there and finally using some method (e.g. MLE) for estimating
the distributions. Probably some simple smoothing method should be used not
to get null probabilities.

A possible method is described in the thesis of Lluís Márquez i Villodre --
"Part-of-speech tagging. A machine learning aproach based on decisión
trees". It would be nice to be able to implement the described tagger using
NLTK classifiers.

Migrated from http://code.google.com/p/nltk/issues/detail?id=475