I've already mentioned this problem in [!314] but I think more
visibility for this problem is useful

The SumfactKernel object stores a tuple of transformations that will be
registered in the realization and applied in the localoperator. These
transformations can change the name of the sum factorization kernel
function. This way it should be possible to use autotuning for picking
transformations.

consecutive means the tail is within the same subelement inames as
the vectorized loop (except for the vectorized iname)
blocked means each tail has its on set of subelement inames