Merge branch 'feature/fix-profiling-fullpath' into 'master'

ref:extensions/dune-codegen using which only works, if generate_operators is
in PATH, otherwise cmake fails. This is especially the case if dune-codegen is
configured completely new.

See merge request [extensions/dune-codegen!317]

  [extensions/dune-codegen!317]: gitlab.dune-project.org/extensions/dune-codegen/merge_requests/317

using which only works, if `generate_operators` is in `PATH`, otherwise cmake fails. This is especially the case if dune-codegen is configured completely new.

Merge branch 'fix/benchmark-opcounting' into 'master'

See merge request [extensions/dune-codegen!315]

  [extensions/dune-codegen!315]: Noneextensions/dune-codegen/merge_requests/315

Merge branch 'feature/donkey-autotune-setup' into 'master'

See merge request [extensions/dune-codegen!313]

  [extensions/dune-codegen!313]: Noneextensions/dune-codegen/merge_requests/313

Merge branch 'feature/blockstructured-vectorization-for-arbitrary-blocksize' into 'master'

ref:extensions/dune-codegen This enables vectorization for arbitrary number of
blocks by manually handling the tail of the vectorized loop. The tail can also
be vectorized, if a smaller vector length fits.

Another approach could be to use padding, but I'm currently not sure how to do
that.

See merge request [extensions/dune-codegen!311]

  [extensions/dune-codegen!311]: gitlab.dune-project.org/extensions/dune-codegen/merge_requests/311

Merge branch 'feature/autotuning-google-benchmark' into 'master'

ref:extensions/dune-codegen Make it possible to use google-benchmark for
autotuning.

See merge request [extensions/dune-codegen!312]

  [extensions/dune-codegen!312]: gitlab.dune-project.org/extensions/dune-codegen/merge_requests/312

Merge branch 'experiments/autotune' into 'master'

ref:extensions/dune-codegen The scaling of the autotune results was bugged.

See merge request [extensions/dune-codegen!310]

  [extensions/dune-codegen!310]: gitlab.dune-project.org/extensions/dune-codegen/merge_requests/310

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

odd behavior for unstructured tensor test...