Make a corpus-specific function for producing a frequency-countable token list

[gregdan3]

The corpus study for ilo Muni would benefit from preserving case, numerals, and potentially emotes. However, using the filtered and cleaned output directly from this library doesn't allow for that since it has already removed everything in the ignorable filters by the time scoring occurs.

In other words, it may be necessary to have a second path out from the main scoring path: before cleaning or filtering, a separate cleaning and filtering takes place which produces a frequency-countable version of the sentence.

• Should preserve case. (already done since case is used to detect names)
• Should preserve numerals. (currently removed during the ignoring filters)
• Should still reduce consecutive duplicates, but not for numerals (oh no)