Cluster Autoscaler 1.12.0

• Fix bug in scaleup related to ready condition handling
• Report reason why pod didn't trigger scale-up
• Fix fetching resource limits on GKE
• Fix GKE operation wait timeout
• Ignore resources with Cloud TPU prefix
• Update the aws ec2 instance types
• Adding the "cluster-autoscaler.kubernetes.io/safe-to-evict": "false"
• annotation to a pod prevents the cluster autoscaler from touching it.
• Add GPU-related scaled_up & scaled_down metrics
• Various fixes

Image: gcr.io/google-containers/cluster-autoscaler:v1.12.0