Stop relying on /mnt/logs for delete_instance.sh

Eventually we should strive to get rid of writing all logs to a centralized persistent volume. This doesn't scale well and this would allow us to simplify things by not having to do this persistent volume claim anymore (which in turn removes the need for the catalogue and scheduler charts to wait on the init chart, shaving ~20 seconds from the deploy time).

At the moment, this is centralized log collection is used in only two places:

• To collect a tar of all the /var/log contents of the pods in delete_instance.sh
• For monitoring

Both of these issues can be easily solved separately. This issue is to address delete_instance.sh. This can be done trivially by using kubectl cp and manually extracting the relevant files from the pod/container