fluentd deployment prevents safe cluster-autoscaler scale in due to pos-files volume

[wendtek]

Due to the use of a hostPath volume for pos-files in the fluentd deployment, cluster-autoscaler is unable to evict pods for safe rescheduling on another node without risk of duplicate ingestion.

I suspect it may be better to use a persistent volume claim that can follow the pod to a new node, though there may other considerations I'm missing.

[pmalek-sumo]

We're not using tail input plugin so this might actually be unused as of now.
I'll take a closer look at it.