fix wrong type indicator(for jpn)

[atuyosi]

I found a few mistake that cause undesirable ocr result in jpn/jpn.unicharambigs.
Some substitution rule do not mandatory (even if vertical writing Japanese).

For example (bad result):
google -> g。。gーe
1985 -> ー985

So, I modified "type indicator(5th column)" .

[tfmorris]

This seems reasonable to me, but how widely tested has it been? ie what is the risk for regressions on other text?

[atuyosi]

I tested this both horizontal writing and vertical writing (of course, Japanese text which contains alphanumeric characters).
I seem that the original jpn.unicharambigs file considers only vertical text layout and some rare cases.