Undefined RTCPeerConnectionState states

[henbos]

There are combinations of transport states that don't apply to any of the states in the table. For example, if all transports are "connecting" except one that is "closed", there is no corresponding RTCPeerConnectionState (because for PC/"connecting" none of the transports can be "closed" and for PC/"closed" the PC has to be closed - what if one of the transports closed due to close() or due to a close_notify alert?).

[alvestrand]

when the remote party sends a DTLS close (which happens when the remote party calls pc.close()), the DTLSTransport will go to "closed". I think that's the only way for a DTLSTransport to get to closed state. It's a final state for the DTLSTransport.

[henbos]

"closed" refers to [[IsClosed]], I think reusing that to mean "the other guy told us they closed" would be a breaking change.

So "failed" or "disconnected" makes more sense.
I propose to make a PR that makes DTLS "closed" cause PC "disconnected" since it's not a failure code it's a close code.

[henbos]

Might also want to see what existing implementations do...

[henbos]

Merging this into #2499