Control messages are messages that are broadcast to all clients and have
no clearly defined content. The idea is that this can be used to
control a monitor without having to keep adding support for specific
commands on the protocol layer.
This also changes some of the existing messages and adds another
ridiculous convenience layer to our HTML/JS templating: data-eval.
We should probably just bite the bullet and use some reactive framework.
In its current state the implementation should allow a user to resume
their session if the websocket connection is reset, for whatever reason.
This could be expanded to allow session sharing (multiple agents logging
in to the same client), or manual session resume via some sort of
password (encode uid & key to some pass-phrase kinda thing, or QR code).
