Qt's Shift-JIS codec behaves different from ICU with regards to the backslash and tilde ascii characters

When using the Shift-JIS codec in Qt to convert to Unicode, the characters 0x5c and 0x7e are mapped to the yen and overline symbols - according to JIS X 2010.

ICU on the other hand maps these characters to backslash and tilde - ascii compatible.

We should adjust Qt to behave like ICU with respect to that.

This issue becomes particularly visible with WebKit, where other WebKit browsers are using ICU. A real-world example is for example the site http://www.bidders.co.jp/ - which specifies shift-jis as encoding and loads jquery via a script tag. At that point it is assumed that jquery is encoded
in shift-jis. JQuery uses regular expressions, which in turn commonly contain backslash characters. With icu these are preserved and the regular expression
parsing proceeds normally. With Qt the backslash is mapped to yen (unicode yen 0x00a5), which breaks parsing.

In addition there's a WebKit layout test that fails due to this difference: https://bugs.webkit.org/show_bug.cgi?id=39852

The attached test-case demonstrates the difference. It can be compiled with g++ -o test main.cpp `pkg-config --cflags --libs QtGui` `icu-config --cflags --ldflags`