feature: OOM / resource starvation / kernel panic resiliency

[azenla]

Currently, krata does not define what happens when OOM situation, general resource starvation, or even kernel panics occur. We should have a clearly defined system for determining this state, and rectifying it. The Zone should have a policy of what to do. For example, some orchestration tools might want the Zone to be destroyed, while some might want it to restart.