HTTP Based Discovery as an alternative to MultiAddress

[luongngocminh]

In the current state, in order to initialize a cluster, a seed in a form of a multiaddr will be required. This seed can only be decided when there're at least 1 node already running (unpredictable port and ip addresses). Due to this limitation, we cannot seamlessly spin up a cluster without having some overhead logic for acquiring the initial seed.
So instead of using multiaddr explicitly, we can opt for an HTTP Based Discovery using API Endpoints somehow.