fix(aws-eks): nodegroup scaling defaults and apply version updates when they settle

[NitinKumar004]

Two EKS fixes (tracker rows EKS-X7 and EKS-X10).

EKS-X7: nodegroup scaling defaults and validation

CreateNodegroup without a scalingConfig used to store 0/0/0, so the nodegroup had no nodes. Real EKS defaults to minSize 1, maxSize 2, desiredSize 2. The provider now applies that default, and with the node sync from desiredSize a default nodegroup brings up 2 Nodes.

Validation now follows the API docs and the messages EKS returns:

• minSize and desiredSize at least 0, maxSize between 1 and 450 (the default "Nodes per managed node group" quota)
• Minimum capacity X can't be greater than desired size Y / desired capacity X can't be greater than max size Y
• CreateNodegroup takes all three sizes or none; a partial scalingConfig is an InvalidParameterException
• UpdateNodegroupConfig with a partial scalingConfig is merged onto the current sizes inside the provider, under its lock, and the merged result is validated. The wire handler no longer does a Describe and merge itself.

Driver change: NodegroupConfig.ScalingConfig is now a pointer (nil means default), and NodegroupConfigUpdate.Scaling is a new NodegroupScalingUpdate with optional sizes.

EKS-X10: version updates apply when the update settles

With --async-settle, UpdateClusterVersion changed the cluster version and the data plane's /version right away, while the update was still in progress. Real EKS keeps reporting the old version until the update completes. Now:

• DescribeCluster reports the old version while the cluster is UPDATING, then the new one once it settles
• /version switches at the same moment. The data plane gets a new SetClusterVersionAt, which holds the current version until a given time on the caller's clock. Snapshots store the target version.
• UpdateNodegroupVersion works the same way: the nodegroup keeps its old version and release version until its update settles
• update records from cluster and nodegroup config/version updates read InProgress in the response and in DescribeUpdate until the change settles, then Successful
• nodegroup and add-on version resolution uses the version the control plane actually runs, so a nodegroup can't move to 1.32 while the cluster is still upgrading to it

With async settle off, all of this stays immediate as before.

Verification

• New tests (FakeClock for settle): nodegroup_scaling_test.go, version_settle_test.go, sdk_nodegroup_scaling_test.go, and TestServerVersion_ScheduledSwitch. Each fails on the old code.
• go build, go vet, and go test -race on providers/aws/eks, server/aws/eks, services/kubernetes and persist; golangci-lint with --new-from-rev shows 0 issues.
• cloudemu serve --async-settle with the aws CLI and kubectl: create-nodegroup with no scaling config gives min 1/max 2/desired 2 and 2 nodes in kubectl get nodes; a partial config is rejected. After update-cluster-version, describe-update is InProgress while describe-cluster shows 1.31 UPDATING and kubectl version shows v1.31.x. Both switch to 1.32 once describe-update reads Successful. update-nodegroup-version behaves the same way.
• Terraform aws_eks_node_group requires scaling_config in the provider, so it was tested with one: apply, plan clean, in-place cluster upgrade 1.31 to 1.32 with desired_size 2 to 3, plan clean, destroy.