Better matching for titles with roman numerals

[jingai]

We currently attempt to swap out digits in replaceDigits()/word_form() when fuzzy matching if we didn't get a good match without it (e.g., five and 5), but roman numerals are very common in movie titles too.

In particular, if all that distinguishes one movie from another is the roman numerals, we're likely to either match the wrong movie or none at all.

Take for instance the user requesting: "Alexa, ask Kodi to watch movie back to the future two"

And in the user's library, we have these titles:

Back to the Future
Back to the Future II
Back to the Future III

The fuzzy match currently will stop on and select "Back to the Future" (the first movie).

[jingai]

And this is even more difficult when we're supporting multiple languages... the existing word_form() method only supports English, too.. probably should fix that at some point.

[jingai]

This is done and pushed up, but as noted, word form number conversions are only supported for English.

Closing the issue as the language issue is arguably separate.