Add gateway/worker protocol to decouple persistent frontend ingress from simulation runtime

[skim0119]

Short description

Introduce a gateway/worker protocol so the frontend-facing VR server can stay up on server-machine while simulation workers run independently on client machine.

The goal is to decouple:

• persistent client ingress and session routing
• simulation execution and scene publishing

This would support a deployment where:

• clients connect only to server machine
• machine A forwards session/input traffic to a worker on client-machine
• machine B sends scene state, asset updates, and errors back through server
• workers can be restarted or swapped without changing frontend deployment (key)

Scope

• define gateway/worker message protocol
• support session setup/teardown, hello, xr_input, heartbeat, reset
• relay scene state and asset/overlay updates
• handle worker registration, liveness, disconnects, and error reporting

Note

• Server can remain continuously available as the stable frontend endpoint
• Client can run and restart simulation workers independently
• client input reaches workers through server
• worker output is relayed back to clients through server