Use other storage instead of etcd for karmada-aggregated-apiserver and karmada-search

[majoyz]

What would you like to be added:
Use other storage(like metrics-server used) instead of etcd for karmada-aggregated-apiserver and karmada-search.

Why is this needed:
Currently, we use etcd as karmada-aggregated-apiserver and karmada-search's storage. If there's no etcd server expose(in some public cloud k8s clusters), we have to deploy an etcd in karmada cluster, is this necessary? Can we just store these resource(clusters and resourceregistry) in memory or other places without third party dependencies?