Tidy-up AssmbScatr operators implementation

This MR is a tidy-up of the AssmbScatr operators. It also makes sure that temporary storage and output are properly zeroed in the AssmbScatrCUDA operator.

changed the description

marked this merge request as ready

requested review from @ccantwel

assigned to @ccantwel

approved this merge request

mentioned in commit 5a7510d2

merged