Add RTCSctpTransport.setTransport

[lgrahl]

These changes add a setTransport to the SCTP transport to allow seemless changing of the underlying DTLS transport. It also adds a missing state check for the DTLS transport instance on construction.

Some questions are left:

1. DTLS transport state on construction/setTransport

When constructing an instance that requires a DTLS transport, the description is:

If an attempt is made to construct an RTCRtpSender object with transport.state or rtcpTransport.state closed

Shouldn't the failed state be covered by the description, too? (The same would need to be applied to the description of setTransport.)

Answer: No.

2. state vs. ObjectNameState

I've seen that in the RTCSctpTransport to reference the current active state, RTCSctpTransportState is being used instead of just state (to refer the local attribute) as in the RTCQuicTransport.start method. Should we unify this for consistency (and by which method)?

3. DTLS transport component check

Do we need a dtlsTransport.transport.component check on construction of the SCTP transport (and on calling setTransport)? If the component is not rtp, this should be an error, right?

[lgrahl]

Also fixes a minor compatibility issue raised in #755.

[taylor-b]

If the component is not rtp, this should be an error, right?

Why? If you're not RTCP-muxing, you need separate DTLS associations for RTP and RTCP.

Somehow missed that we're talking about SctpTransport, not RtpSender. Nevermind, you're correct.

[lgrahl]

@aboba @robin-raymond Any comments to 3? Do you agree with @taylor-b and me?

[lgrahl]

Any action needed from my side to get this merged (other than the conflict that appeared in the meantime 🙂)?

[lgrahl]

Ping.