cell->super

I thought we fixed this a few weeks ago but no we still have a problem with the setting of the cell's super pointer.

As it is now, scheduler_reweight() and other functions will crash if cell->super is not set. So we can't set a super pointer and a g_super pointer. We need to have only one of those correctly set in all three cases:

• only hydro tasks
• only gravity tasks
• both hydro and gravity If we can handle this in a generic fashion then adding more particles in the future will be easier.

So the questions are:

• how do we set super in all three cases ?
• how do we deal with cases where the hydro and gravity tasks are attached at a different level in the tree. Do we allow that ?
• Should we separate the construction of the ghost/kick/init/... tasks from the actual setting of the super pointer ?

I think the answer to the last question is yes and that we should not allow for tasks to be attached at different levels in the tree.