Handle new-initiator/new-responder correctly

[lgrahl]

When a 'new-initiator' or 'new-responder' message has been received, the initiator/responder simply replaces the corresponding instance but does not reset its own state.

Due to a limitation of the implementation, after the task has kicked in, it should also...

• drop the responder in case it is an initiator, and
• close the connection normally in case it is a responder.

[dbrgn]

Good point. I assume with "its own state" you mean some fields in the Common struct.

If I'm not mistaken, if the task has not yet kicked in, we only need to reset self.common.signaling_state.