QCborValue: fix double-accounting of the usedData when decoding strings

We can only update usedData at the end, after we've decoded all chunks.
The update inside the lambda was double-accounting for the first chunk,
which could lead to signed integer overflows in usedData.

Not unit-testable since the usedData value is not visible in the API.

Change-Id: Ibdc95e9af7bd456a94ecfffd16061cc955208859
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>

src/corelib/serialization/qcborvalue.cpp

@@ -1561,8 +1561,6 @@ void QCborContainerPrivate::decodeStringFromCbor(QCborStreamReader &reader)
         if (newSize > MaxByteArraySize)
             return -1;
 
-        // since usedData <= data.size(), this can't overflow
-        usedData += increment;
         data.resize(newSize);
         return offset;
     };