utf8 to utf16 conversion must use native endianness

There is no obvious reason why native endianness should not be used, and
tests failed on big-endian architectures.

Author: tautschnig

jbmc/src/java_bytecode/java_string_literals.cpp

@@ -106,7 +106,7 @@ symbol_exprt get_or_create_string_literal_symbol(
   if(string_refinement_enabled)
   {
     const array_exprt data =
-      utf16_to_array(utf8_to_utf16_little_endian(id2string(value)));
+      utf16_to_array(utf8_to_utf16(id2string(value), false));
 
     struct_exprt literal_init(new_symbol.type);
     literal_init.operands().resize(jls_struct.components().size());

src/util/unicode.h

@@ -24,6 +24,7 @@ std::wstring widen(const std::string &s);
 
 std::string utf32_to_utf8(const std::basic_string<unsigned int> &s);
 
+std::wstring utf8_to_utf16(const std::string &in, bool swap_bytes);
 std::wstring utf8_to_utf16_big_endian(const std::string &);
 std::wstring utf8_to_utf16_little_endian(const std::string &);
 std::string utf16_little_endian_to_java(const wchar_t ch);