fix: address AKS ci flakiness

.github/test-infra/azure/aks/main.tf

@@ -115,4 +115,5 @@ resource "azurerm_kubernetes_cluster_node_pool" "worker" {
   auto_scaling_enabled  = var.enable_autoscaling
   min_count             = var.enable_autoscaling ? var.autoscaling_min_node_count_worker : null
   max_count             = var.enable_autoscaling ? var.autoscaling_max_node_count_worker : null
+  node_count            = var.worker_node_pool_count
 }

.github/test-infra/azure/aks/variables.tf

@@ -76,6 +76,12 @@ variable "default_node_pool_vm_size" {
   type        = string
 }
 
+variable "worker_node_pool_count" {
+  description = "Number of nodes to add to the worker nodepool"
+  default     = 3
+  type        = number
+}
+
 variable "worker_pool_vm_size" {
   description = "Specifies the vm size of the worker node pool"
   default     = "Standard_F8s_v2"