Typha autoscaler should exclude Fargate Nodes in node count

[tscn]

Expected Behavior

As Fargate nodes cannot run calico-node, they should not be included in the node count that is used for Typha autoscaling.
Seems there is logic for a similar thing in AKS in place: https://github.com/tigera/operator/blob/master/pkg/controller/installation/typha_autoscaler.go#L243

Current Behavior

Typha autoscaler includes Fargate nodes when calculating schedulable nodes.

Possible Solution

Exclude nodes with label eks.amazonaws.com/compute-type=fargate in func (t *typhaAutoscaler) getNodeCounts() (int, int, error)

Context

I currently have 2 Fargate pods/nodes in my cluster that are running Karpenter and once Karpenter will scale out and add 3
additional nodes to the cluster, typha will scale to 3 instances and prevent a consolidation to less instances.

Your Environment

• CSP: AWS
• Service: EKS
• K8s version: 1.24
• Tigera Operator: v1.29.0