Add support for terminating multiple pods within a topology

[linki]

In order to simulate an outage of an entire node or availability zone, we could add some kind of selection based on topology keys.

For instance, starting chaoskube with

--topology-key=kubernetes.io/zone

Could batch-kill all pods within a single zone per iteration.

Similarly,

--topology-key=kubernetes.io/node
--topology-key=kubernetes.io/region

would batch-kill all Pods of a single node or region, respectively.

Combined with the already existing --max-kill flag (default 1) one could trigger deletions for multiple nodes, zones etc.

Furthermore,

--topology-key=kubernetes.io/pod

would be equivalent to the current behaviour, I believe.

Pods would probably restart on the same nodes after being terminated but it might still be useful to simulate full node or zone outages.