Fixes "skip over" invalid character sequence code inside #else part of
#ifdef HAVE_GNU_LIBICONV. Looks like a typo.

Append "//TRANSLIT" to iconv conversion, so best effort replacement
characters will be used when no direct character representation exist.

--- lib/Unicode.cc.orig	Mon Jan  3 01:42:53 2005
+++ lib/Unicode.cc	Tue Jul  9 12:08:50 2013
@@ -125,7 +125,7 @@ namespace bt {
             // POSIX compliant iconv(3)
             inp =
               reinterpret_cast<char *>
-              (const_cast<typename _Source::value_type *>(in.data()));
+              (const_cast<typename _Source::value_type *>(in.data())) + off;
 #endif
             in_bytes = in_size - off;
             break;
@@ -168,6 +168,7 @@ bool bt::hasUnicode() {
 
 #ifdef HAVE_NL_LANGINFO
   codeset = nl_langinfo(CODESET);
+  codeset += "//TRANSLIT";
 #else
   std::string locale = setlocale(LC_CTYPE, 0);
   std::string::const_iterator it = locale.begin();