Integrate multiple inference node selection based on the node's public information

[jorgeantonio21]

Feature Request

• Various nodes will be providing inference compute to the Atoma Network (that should be leveraged by Utopia). The nodes will share their public IPs, for the time being, and we should store these on the openweb ui for later retrieval.
• When a new request user request comes in, we can select a node to process its chat requests. This node can be selected randomly for now, and be kept available for the user for some pre-defined time (say a day).
• We need to make sure that if the node is unresponsive, to pick another one to serve the user (and update the node assignment for the specific user).