WIP: fix #546: Better word wrap for plain text

[bobbylight]

Proposed fix for #546. Use BreakIterator in PlainTextTokenMaker to ensure line wrapping is done in a per-locale, language-specific basis.

Remaining issues:

• Mark Occurrences marks tokens as identified, which isn't logical to users (tokens area identified based on where line breaks can occur, so stuff like foo). is marked as a single token). We'll need to either disable Mark Occurrences for plain text, or create a custom PlainTextOccurrenceMarker that looks for e.g. strings of letters within tokens (which might be very very slow).
• Would need to delete PlainTextTokenMaker.flex, as PlainTextTokenMaker.java is now a hand-written TokenMaker.