diacritic opt-in for frequencies

[paceaux]

right now n-grams work by stripping diacritics; all frequencies/ results take away accent marks.

We should at least offer the ability to opt-in to diacritics.
Seeing confusing results on the demo:

in portuguese, ão is listed, but we can't highlight (a problem with the demo, but still)

With French, we have éèê to account for and it may be useful to see this separate.

[paceaux]

This should probably be a few pieces:

1. There should be an ability to identify characters with diacritics and keep them in a separate place
2. There should be an ability to match characters with diacritics to ones without (i.e. knowing that èéê are all still e)
3. Frequency analysis with diacritics on/off.