It does not yet do all the jobs it can do, but it exists and we can
run tests with it.

by introducing an order on the kernel objects and
using the first n kernels instead of an itertools.permutations
generator

We will need even more "levels" in the near future

So far, it did not set the qp tuple correctly and
thus sometimes resulted in too big numbers being
chosen.

Instationary stuff not fixed so far.

Now working apart from iteration order deterministicness

will be superseded by our actual model anyway.

Still not beautiful...

It always aims for maximally horizontal vectorization
where possible.

By using loopys group mechanism. Each sum factorization kernel
defines a group that conflicts with all other sum factorization
groups.

Conflicts:
	python/dune/perftool/sumfact/realization.py
	python/dune/perftool/sumfact/vectorization.py

Save all stage 1 sum factorization kernels that are used in
accumulation expression in the cache during the dry run. Discard all
inactive sum factorization kernels in decide_vetorization_strategy.