density and bulk velocity unnecessary ghost filling?

[nicolasaunai]

Currently, we fill total ion moment ghosts after they have been computed.
The reason for this is that, although currently numerical calculations do not need moment ghosts, coarsening stencil for primal takes 3 fine primal nodes (0.25, 0.5, 0.25), which, for coarse nodes falling on top of fine patch level border, will hit its level ghost nodes.

There are two issues for that:

• the current schedules overwrite the interior border (on refinement only at the time this issue is written since overwrite_interior=false only counts for same level overlaps (in calculateOverlap()). This will not be an issue after fine_boundary_represents_var for copy **before** refinement #1062 is merged (or another taking that code) where this boolean also counts for refinement (in computeFillBoxesOverlaps).

• it's possibly waste of time for no gain because it would not be needed if the coarsening of the moments was done by only taking the primal fine value on top of the coarse fine one. This would not need the ghost.