Fix cluster startup errors

[phudtran]

This PR fixes the following issues:

• Runtime errors where a dictionary changed size while being iterated through.
• An issue where the veth creation returned before the device was created.
• Timeout errors caused by the mizar-daemon taking longer than 15 minutes to come up in the Arktos kube-up environment.
• A runtime error where delete was called on a nonexistent key in a dictionary.
• An issue where pods deployed before mizar did not have network connectivity