Add an segmenter option whether we use dictionary (support u-dx)

[makotokato]

This is a low priority issue from https://bugzilla.mozilla.org/show_bug.cgi?id=1871754. Before using ICU4X, Gecko's word segmenter for Chinese and Japanese is that segment is whether character class is same or not.

Actually, word segmenter for Chinese and Japanese are based on dictionary. Since new words are always incremented, dictionary implementation may not be enough for quality without updating it.

Although we are considering to use other ways for it such as Machine Leaning in the future, it may be better that we have a segmenter's options not to use dictionary for some languages only (If Japanese, we don't use dictionary, but other can use it).

CC: @aethanyc