Controller: SCO implementation

[zxzxwu]

This provides a capability to test SCO virtually. We will also apply the same strategy to CIS.

[zxzxwu]

@barbibulle I suggest we should abandon the support of RemoteLink, so that the type annotation between controller and link would be simpler.

[barbibulle]

@barbibulle I suggest we should abandon the support of RemoteLink, so that the type annotation between controller and link would be simpler.

I agree. RemoteLink served its purpose in the past, but there are easier ways to achieve the same with a slightly different topology: instead of having a local controller that can exchange data with a remote controller via a remote link relay, the alternative is to have a relay that exposes HCI interfaces over a remotable transport (TCP, WebSocket) rather than "link" interfaces. We can now use RootCanal as such a remote HCI hub.