fix: add speaker counterpart information to transcription

[Stell0]

Summary

This PR introduces the counterpart speaker information in transcriptions and prevents the connection from being blocked during the summary operation.

Changes

• Transcription output

  • Added speaker_counterpart_name and speaker_counterpart_number fields to each transcription segment.

• Connection handling

  • Fixed an issue where the connection could be blocked during the summary step.

Impact

• Extends the transcription data structure (backward compatible, but clients should handle the new field).
• Improves stability and prevents potential blocking during summary execution.