fix: support pyannote 4 diarization API

[FernandoAbishai]

Context

Phase 5B.6 real-media qualification exposed a pyannote API compatibility
issue in the existing diarization adapter.

The repository permits pyannote.audio 4.x, while the adapter still assumed
the pyannote 3.x authentication and output APIs.

Changes

• support both token= and legacy use_auth_token= authentication APIs;
• normalize pyannote 3 Annotation and pyannote 4
  DiarizeOutput.speaker_diarization results;
• preserve existing overlap-based word/segment speaker assignment;
• preserve graceful no-token/model-load/runtime failure behavior;
• redact HF tokens from logs;
• add focused compatibility smokes.

Validation

• focused diarization tests: 7/7 PASS
• backend smoke suite: 100/100 PASS
• optional-capability smokes: 8/8 PASS
• python compileall: PASS
• qualification harness: PASS
• git diff --check: PASS

Scope

No changes to:

• requirements
• qualification matrix
• transcription engines
• clip discovery
• review
• captions
• export
• project schema
• Electron/runtime
• release workflows

This is the narrow P1 corrective fix discovered during Phase 5B.6.

Do not merge.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.