[agents][incident flup] wait for server side agent to join remove before snending data

[AngeloGiacco]

i think we can add this for extra safety as a flup to the incident, logic looks reasonable to me. could you take it for a spin locally and see if it works well @kraenhansen ?

Note

Medium Risk
Adds a new pre-send wait/timeout in the WebRTC connection handshake, which could delay or fail connection setup if participant events/identities differ from expectations.

Overview
Prevents WebRTC conversation init messages from being dropped by waiting for a remote participant whose identity contains agent to appear (either already in room.remoteParticipants or via RoomEvent.ParticipantConnected) before sending the initiation/overrides payload, with a 30s timeout.

Updates the LiveKit mocks in WebRTCConnection.test.ts to include remoteParticipants and the ParticipantConnected event so tests align with the new handshake behavior.

^{Written by Cursor Bugbot for commit 76bb8ce. This will update automatically on new commits. Configure here.}

[kraenhansen]

I've been unable to reproduce the original issue, so I can't reliably determine if this will fix it. To clarify: You mostly just want me to validate that this doesn't break stuff?

As I see it this change will add the agent's connection to the critical path of starting a session, right? That's not ideal and should be avoided if we'd want to keep things fast and responsive 😬

[AngeloGiacco]

would be great yeah :) i dont have rn set up so would like to check it doesnt cause any issues there but otherwise i will get round to it later this week, can always just QA on js otherwise

[AngeloGiacco]

btw i dont think this would have actually prevented the incident, it was definitely the livekit bug, just looks safer to me