Inconsitent edge orientations in t8_eclass.c

[jfussbro]

Currently the lookup table t8_edge_vertex_to_tree_vertex stores the vertices of an edge in a way, that the smaller vertex number is the first and the higher one comes second. In the case of the triangle we therefore get: { { 1, 2 }, { 0, 2 }, { 0, 1 } }, /* triangle */ In order to keep the orientation of the edges in a closed loop, the order should be changed to: { { 1, 2 }, { 2, 0 }, { 0, 1 } }, /* triangle */.

The same also applies for tetrahedral and for prism elements.

[lukasdreyer]

This relates to https://github.com/DLR-AMR/t8code/issues/774 in that there currently is no clearly defined reference element for each shape, that is mapped to the refined element in an orientation preserving way, and whose edge/face/vertex enumeration is used for enumeration of the refined element. The enumeration used for the default scheme enumerates the vertices of the subfaces (here edges) in a lexicographical way. The information regarding reference elements in this file should be collected in a new file and documented, and we should have a discussion, whether to use orientation preserving enumeration or the lexicographical enumeration, that ensures that two neigbouring faces have the same orientation