There was a mismatch in the condition. We don't want to copy scalars only when
the filter added new scalars for the input. Fixed that check. Otherwise, when we
tried to clip with point data scalars for an AMR dataset in ParaView, the filter
would croak and segfault.