Get rid of new_blah_hacks; diagflat batch rule

Author: zou3519

functorch/csrc/BatchRulesStopDecomposition.cpp

@@ -196,7 +196,6 @@ TORCH_LIBRARY_IMPL(aten, FT_BATCHED_KEY, m) {
   STOP_DECOMPOSE(cumsum.dimname_out);
   STOP_DECOMPOSE(cumsum_.dimname);
   STOP_DECOMPOSE(data);
-  STOP_DECOMPOSE(diagflat);
   STOP_DECOMPOSE(diagonal.Dimname);
   STOP_DECOMPOSE(diff.out);
   STOP_DECOMPOSE(divide.out);