_PyUnicode_Writer() now also reuses Unicode singletons:

empty string and latin1 single character

Objects/unicodeobject.c

@@ -12924,7 +12924,7 @@ _PyUnicodeWriter_Finish(_PyUnicodeWriter *writer)
         writer->buffer = newbuffer;
     }
     assert(_PyUnicode_CheckConsistency(writer->buffer, 1));
-    return writer->buffer;
+    return unicode_result_ready(writer->buffer);
 }
 
 void