segmentForDiff() is supposed to allow character-level diffing of Chinese
text, by adding spaces and then removing them after the diff is
complete. But when I tested it for I2d0a6996b02d37a3, unsegmentForDiff()
failed to remove the spaces, since there was an <ins> tag between the space
and the Chinese character.
So instead, use formfeed characters to separate the Chinese characters,
and strip them unconditionally instead of relying on them being next to
Chinese characters.
Add test.
Change-Id: I230d8261bbda34ad313785a1f7c31d4db7bf989b
