When user wants to do manual scale-up of ASG by increasing the
desired_capacity and also update some member properties as
part of a stack update, if it fails due to insufficient
resources when resizing, trying to manually scale-down by
decreasing the desired_capacity won't work, as it would
always expect to update the group with the earlier size
before scaling down.
This patch uses the target_size when building the batches.
(cherry picked from commit