Fragile relation between slices added in partitioner and slice number used in laik_map

[weidendo]

Currently, LAIK LULESH seems to assume that the ordering of slices as they added to the partitioning in the partitioner algorithm defines the slice numbers.

However, slice numbers are just meant to be local indexes in an iteration loop when the application wants to go through all slices, to just be consumed locally be laik_my_slice and laik_map.

So the current code may work at the moment, but it is very fragile and may get wrong when changing small implementation details in LAIK (such as not using a stable sorting algorithm internally when post-processing the slices added by a partitioner).

In addition, slight changes in the partitioner result in all laik_map calls to become wrong.

Solutions:

• ask LAIK to return the correct slice, depending on a global index in the index space
• use the introduced tag numbers to identify the same slice/slice group