Correct handling of periodicity

[jbigot]

See https://gitlab.maisondelasimulation.fr/gysela-developpers/voicexx/-/issues/20

[jbigot]

From my point of view, we need to add a notion of local elements for that:

• a subset of the global elements is chosen,
• a first set of local elements is defined as the "primary" local elements, they are in bijection with these global elements,
• additional "non-primary" local elements can be added, a mapping to the global elements is defined, so that at the end the full set of local elements is mapped in a non-injective way to the global elements subset,
• a total order is defined on the local elements.

Chunks would be defined over ranges of local elements instead of global ones.

In case

1. the global element subset is an interval according to the global element ordering,
2. only primary local elements are selected,
3. the order chosen over global element is that of the associated global elements,

then we fall back to the current situation.

This would allow us to define a local element range:

[ -2:5, -1:6, 0, 1, 2, 3, 4, 5, 6, 7:0, 8:1 ]

where l:g means that local element l is non-primary and mapped to global element g, and e means that the primary element e is mapped to global element e.