radio: diarization pipeline fixes, benchmark setup, test episode set

- Fix voice_profiler threshold bug (HOST label overwrote Unknown unconditionally)
- Audio preload optimization: single ffmpeg per episode, 149.5x realtime on 5070 Ti
- WavLM threshold raised to 0.85 (Mike 0.90-0.99, callers 0.46-0.83)
- Promo/bumper filter: weighted signature scoring, 42->27 clean Q&A pairs
- Text-only Q&A fallback for episodes with no CALLER diarization labels
- TRANSFORMERS_OFFLINE=1 to skip HuggingFace freshness checks
- Add diarize_2018.py for targeted re-run + FTS5 rebuild
- Add benchmark.py + BENCH_SETUP.md for GURU-BEAST-ROG (RTX 4090) comparison
- Commit 9-episode training diarization.json outputs
- Session log: 2026-04-27-diarization-pipeline.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

projects/radio-show/audio-processor/training-data/transcripts/2011-09-10-hr1/diarization.json

@@ -0,0 +1,39 @@
+{
+  "num_speakers": 2,
+  "speaker_map": {
+    "HOST": "HOST",
+    "CALLER": "CALLER"
+  },
+  "turns": [
+    {
+      "speaker": "HOST",
+      "start": 0.0,
+      "end": 20.0,
+      "confidence": 0.9
+    },
+    {
+      "speaker": "CALLER",
+      "start": 15.0,
+      "end": 25.0,
+      "confidence": 0.83
+    },
+    {
+      "speaker": "HOST",
+      "start": 20.0,
+      "end": 1855.0,
+      "confidence": 0.86
+    },
+    {
+      "speaker": "CALLER",
+      "start": 1850.0,
+      "end": 1860.0,
+      "confidence": 0.78
+    },
+    {
+      "speaker": "HOST",
+      "start": 1855.0,
+      "end": 2505.0,
+      "confidence": 0.93
+    }
+  ]
+}