[Provisioner] Patch when head node not found for launching on existing cluster

[cblmemo]

In RunPod and Lambda (#3865), we directly raise an error when we cannot found head node for a cluster:

skypilot/sky/provision/runpod/instance.py

Lines 68 to 70 in a4e2fcd

	if head_instance_id is None:
	raise RuntimeError(
	f'Cluster {cluster_name_on_cloud} has no head node.')

This is possible when the user manually changes instance's name on the cloud console.

Instead of error out, we can patch one of the nodes to head node. However, due to different runtime that has been installed to the head node, we need to identify the original head node.