Fix Probability Distribution with Limited Character Set

[tallpauley]

If we have a tiny character set like hambugers and ask for words we get too many short words. This is because we've limited the word-wordcount pairs severely, and we're left with very high probability words like a and as, and very low probability words that have the required characters but will hardly show up:

wsv = wordsiv.WordSiv(limit_glyphs='hambugers')
wsv.words(lc=True, source='en_wordcount_web', num_words=10)
['guess', 'as', 'as', 'err', 'a', 'us', 'ease', 'message', 'as', 'mer']

We have the prob parameter (as seen below), but it would be ideal if we somehow can notice this happening and automagically pull from less frequently occurring words.

wsv.words(lc=True, source='en_wordcount_web', num_words=10, prob=False)
['resumes',
 'mrsa',
 'meager',
 'barber',
 'srgb',
 'basu',
 'sag',
 'hab',
 'aea',
 'ssh']