LiveKit video should be turned only on demand

[koschke]

Currently, LivekitVideoManager.Start() will try to obtain a token from the token server even though no user has actually turned on the face cam. As a consequence, the user will be annoyed by an error notification at the start of every play mode if the token server has not been installed and started.
In a development environment, we will rarely install and start this server. Likewise, we should neither assume that "real" users install and start the server.

Rather than trying to obtain the token on Start, we should rather try that only if the user explicitly demands it (starts the face cam).