feat: Kokoro voice blending — blend builder UI + weighted tensor synthesis

Add voice blend support to TTS pipeline and settings UI. Users can mix
2–5 Kokoro voices with per-voice weight sliders; the blended style tensor
replaces the single voice when enabled. Settings persist as JSON and auto-
load on synthesis when no explicit voice is supplied in the request.

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

src/fabledassistant/routes/voice.py

@@ -126,8 +126,27 @@ async def synthesise_speech():
     except (TypeError, ValueError):
         speed = 1.0
 
+    voice_blend = data.get("voice_blend")
+    if not isinstance(voice_blend, list) or len(voice_blend) < 2:
+        voice_blend = None
+
+    # Auto-load blend from user settings when no explicit voice/blend provided
+    if voice_blend is None and "voice" not in data and "voice_blend" not in data:
+        from fabledassistant.services.settings import get_setting
+        from fabledassistant.auth import get_current_user_id
+        import json as _json
+        try:
+            uid = get_current_user_id()
+            stored = await get_setting(uid, "voice_tts_blend", "")
+            if stored:
+                parsed = _json.loads(stored)
+                if isinstance(parsed, list) and len(parsed) >= 2:
+                    voice_blend = parsed
+        except Exception:
+            pass
+
     try:
-        wav_bytes = await synthesise(text, voice=voice, speed=speed)
+        wav_bytes = await synthesise(text, voice=voice, speed=speed, voice_blend=voice_blend)
     except Exception:
         logger.exception("TTS synthesis failed")
         return jsonify({"error": "Synthesis failed"}), 500