fix: 修复 grok-imagine-0.9 同时支持文生图和图生视频

- 恢复 is_video_model: True 以支持视频生成
- 更新模型描述说明支持两种模式
- 保留空列表安全访问修复

根据 xAI 官方文档，grok-imagine-0.9 支持：
1. 文本 → 图片 (text-to-image)
2. 图片 + 文本 → 视频 (image-to-video)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

app/models/grok_models.py

@@ -103,13 +103,13 @@ _MODEL_CONFIG: Dict[str, Dict[str, Any]] = {
         "cost": {"type": "low_cost", "multiplier": 1, "description": "计1次调用"},
         "requires_super": False,
         "display_name": "Grok Imagine 0.9",
-        "description": "Image generation model powered by Grok",
+        "description": "Image and video generation model. Supports text-to-image and image-to-video generation.",
         "raw_model_path": "xai/grok-imagine-0.9",
         "default_temperature": 1.0,
         "default_max_output_tokens": 8192,
         "supported_max_output_tokens": 131072,
         "default_top_p": 0.95,
-        "is_video_model": False
+        "is_video_model": True
     }
 }