Dynamic and persistent cluster-wide ep-engine threads

Currently people modify ep-engine's thread counts using something along the following lines:

curl -u Administrator:password -XPOST -d 'ns_bucket:update_bucket_props("default", [{extra_config_string, "max_num_nonio=8"}]).' http://localhost:8091/diag/eval

The memcached process will honour this setting upon the next restart.

MB-22041 has made ep-engine's threads (readers / writers / auxIO / nonIO) properly dynamic. i.e. the following command actually changes the number of nonIO threads:
cbepctl localhost:11210 set flush_param max_num_nonio 8
However this only applies the change on an individual node and it doesn't persist so must be reapplied after a restart.

This ticket is to unify these two mechanisms via the cluster manager so that users can affect the thread counts dynamically across the cluster and have this setting persist across restarts.