We should hardcode the unicode version so that the expected unicode version is also version controlled and things don't accidentally change under the hood.

See https://github.com/unicode-rs/unicode-segmentation/blob/master/scripts/unicode.py#L57