[boost] Some library in Boost can not build with msvc if system default codepage is multi-bytes charset.

For example, Boost.Regex use some characters with 128-255 ansi code in source file, and msvc will convert them to Unicode using default codepage. If the default codepage is an multi-bytes charset, that convertion would truncate some of string declaration. For Unicode string, sometimes it would not raise an error on compile time, but it does the worse thing. L"aa汉aa" has different means for different codepages, and L"\xAB\x9A" does not. I think all file in boost should only contains 0-127 ansi code, and any other characters need for string or char declaration should use \x instead.