fix: Increase VLM max_tokens to 2000 to avoid response truncation

pixelle_video/services/quality/character_analyzer.py

@@ -167,7 +167,7 @@ Output ONLY the JSON object, no additional text."""
                 model=vision_model,
                 messages=messages,
                 temperature=0.3,
-                max_tokens=800
+                max_tokens=2000  # Increased to avoid truncation
             )
             
             vlm_response = response.choices[0].message.content if response.choices else None