Stream directly from Hubs Cloud rooms to services configured in the admin panel

Nobody has claimed this yet.

As we ponder doing streaming for conferences, etc., from our custom hubs install, it occurs to me that key feature that would make this so much easier would be a "stream camera" in hubs.

If we entered the service keys in the Admin panel, Hubs could set up a custom version of the client that ran in an appropriate AWS instance, and rendered out to a stream. Being able to create a "video camera" (akin to the current selfie camera) as an admin, and start streaming whatever it is seeing, would be very useful.

Key features:

* don't need to have anyone else there with you. Could set up a "studio" with a camera in place, and just start it up. Anything it sees goes out.
* could tie (somehow) the camera to an avatar (like the current camera view) so it could stream out.
* only admins can create an deploy cameras. Since each camera implies a "not free" rendering instance, that would be key. Since you're only streaming to streams you create, cameras would be tied to streams, and so if you had one streaming already and tried to start a duplicate, you'd get a warning asking if you want to take it over, or not stream.

There are a lot of other things that could be done here (e.g., all the OBS scene/transition stuff, so you could "keep the stream on" while moving cameras, switching scenes, switching between multiple camera views with/without transitions, setting up for the next shot, etc).

Our entire VR2020 conference would be an order of magnitude easier if we had this. And for smaller things, this becomes a one-stop-shop for social experience streams and events. Even with current limits, getting all 15-20 speakers for an event into a room would be quite cool.

┆Issue is synchronized with this [Jira Task](https://mozilla-hub.atlassian.net/browse/HUBS-1026)

Start by reviewing the admin panel, custom Hubs client, and AWS rendering-instance approach described in the issue; no files or tests are named. Define the camera and stream lifecycle, admin-only deployment, service configuration, and duplicate-stream behavior, then validate the design against the broader streaming and scene requirements.

Written by the indexing model from the issue text.