Use iterators instead of pulling the whole CHANGELOG into memory

[movermeyer]

Ref: https://github.com/mc706/changelog-cli/pull/21#discussion_r563222258

[mc706]

Started down this path on the JSON export piece. Creating a new "get_changes" util that returns a stream of the file lines instead of the whole file, that way i can use map and filter efficiently to get only the lines i need for export. This will also help on being able to just fetch the file and getting things like "give me the first changeset" using next().