fix: UTF-16 surrogate handling.

The data pointer needs to move before decoding the second
surrogate, and not after. The first surrogate was begin decoded
again, resulting in an invalid codepoint.

This affected clipboard operations originating on Windows machines,
where the text is encoded in UTF-16 and copying characters from a
high plane (like emojis) was broken.

src/lib/base/Unicode.cpp

@@ -459,9 +459,9 @@ String Unicode::doUTF16ToUTF8(const UInt8 *data, UInt32 n, bool *errors)
       setError(errors);
       toUTF8(dst, s_replacement, NULL);
     } else if (c >= 0x0000d800 && c <= 0x0000dbff) {
-      UInt32 c2 = decode16(data, byteSwapped);
       data += 2;
       --n;
+      UInt32 c2 = decode16(data, byteSwapped);
       if (c2 < 0x0000dc00 || c2 > 0x0000dfff) {
         // error -- [d800,dbff] not followed by [dc00,dfff]
         setError(errors);