Fixed another obscure bug that was causing obscure core dumps.
cell_array, used by sprintfs was much too small, so I expanded it. I
do give a post traversal check to see if its expanded too much.
In addition, I'm now doing more resetting when we abort for a tabled
subgoal that is too big. I've also included XSB throws in Ctrace, at
least when they occur from the C level.