[skip ci][WIP] Apply trasformation

Decision about loop order should be done with autotuning in the end.

python/dune/codegen/sumfact/realization.py

@@ -278,6 +278,10 @@ def realize_sumfact_kernel_function(sf):
                                               )
                                   })
 
+    # Register kernel transformations
+    from dune.codegen.sumfact.transformations import reorder_loops_in_tensor_contraction
+    transform(reorder_loops_in_tensor_contraction, 'lkji')
+
     # Construct a loopy kernel object
     from dune.codegen.pdelab.localoperator import extract_kernel_from_cache
     args = ("const char* buffer0", "const char* buffer1") + sf.interface.signature_args