Move DisableScaleDownForLoop to the scaleUp loop

When running the autoscaler with `--max-nodes-total` set and that max is reached downscale is disabled when new unschedulable pods are added.
Preventing the removal of nodes.
If pods are created very frequently by let's say a cronjob, the cluster will never be able to downscale.
In order to fix that behavior i moved the downscale lock from the process to the actual place where scale-up happens.

cluster-autoscaler/core/filter_out_schedulable.go

@@ -77,7 +77,6 @@ func (p *filterOutSchedulablePodListProcessor) Process(
 
 	if len(unschedulablePodsToHelp) != len(unschedulablePods) {
 		klog.V(2).Info("Schedulable pods present")
-		context.ProcessorCallbacks.DisableScaleDownForLoop()
 	} else {
 		klog.V(4).Info("No schedulable pods")
 	}