Ignore Roman numbers

[Mte90]

Right now roman numbers are a problem because they cannot be interpreted, basically they are like symbols but composited.

Ideally is better to remove stuff like X, XIII, MCXX and so on.
Example list at https://gist.github.com/riverrun/ac91218bb1678b857c12#file-roman_numerals-py-L14

So we can append to the blacklist generated this stuff.

[dabinat]

This is an issue I ran into when cleaning up the English wiki. You can't filter out the letter I in English because "I" is a valid word that appears in a large number of sentences. Additionally, you cannot guarantee that Roman numerals will always be capitalized. There are lots of languages with single-character words that could potentially be affected.

However, I can see value in adding an option to filter out all caps words greater than x characters, which would catch at least some of them and also acronyms.

[Mte90]

Yes I think that from II and so on will be not a problem, at least we minimize the issue itself.

[Mte90]

I was thinking that is something that is better to manage on the wiki-scraper or we need to add manually those number on the word_usage file that I think that is not the best case.