Improve the Java variable names corresponding to COBOL variables which contain Shift_JIS characters

[yutaro-sakamoto]

My pull request #324 enables cobj to convert COBOL variable names which contain Shift_JIS characters to valid Java variable names. However, the generated Java variable names are not readable because multiple byte characters are converted to hex decimal forms. For instance, cobj converts X項目 whose byte sequence is 0x58, 0x8d 0x80, 0x96, 0xda to f_588d8096da. I should improve this conversion rule.

Since Java allows us to use multiple byte characters in source code, X項目 can be converted to f_X項目. However it seems that Java does not accept some Shift_JIS characters such as Ｏ－文字列. This is why I apply the current conversion rule.

In order to resolve this issue, it is necessary to identify the set of Shift_JIS characters which Java compiler accepts.

[yutaro-sakamoto]

error_bytes.txt
This file indicates byte sequences of all prohibited Shift_JIS charactes for Java variables

[yutaro-sakamoto]

#501 resolves this issue partially.