fix(transport-webrtc): fix EOF race in readCandidatesUntilConnected

[paschal533]

Problem

A race condition caused the WebRTC "simple" transport compliance test to fail intermittently: the listener closes its signaling stream just before the initiator's connectionstatechange: connected event fires. The initiator then received EOF on the signaling stream and threw an UnexpectedEOFError, discarding a perfectly good peer connection.

Root cause: connectedPromise was declared inside the try block, making it inaccessible in the catch handler for stream read errors.

Solution

Move connectedPromise outside the try block. In the catch handler, if the peer connection is not yet 'connected', await connectedPromise:

• If it resolves → the connection succeeded, ignore the stream error
• If it rejects → the connection failed, re-throw the original stream error

Also corrects the log message from "error parsing ICE candidate" to "error reading ICE candidates" for accuracy.

Test plan

• Added readCandidatesUntilConnected describe block with 4 tests
• Test: throws ConnectionFailedError when the peer connection enters 'failed' state
• Test: does not fail when peer connection is temporarily disconnected then recovers
• Test: throws ConnectionFailedError when peer connection goes disconnected then failed
• Test: returns without error when peer connection reaches 'connected' state

Related PRs

This is one of four focused fixes split from #3406 as requested by @dozyio:

• fix(transport-webrtc): check initial peer connection state at construction time
• This PR — UnexpectedEOFError race in readCandidatesUntilConnected
• fix(transport-webrtc): do not treat 'disconnected' as a terminal peer connection state
• fix(interface-compliance-tests): increase abort propagation timeout in stream muxer close test