Add vision model support and related configurations in Assistant

- Introduce new fields in AssistantConfig, schemas, and database models to support vision capabilities, including `vision_enabled` and `vision_model_resource_id`.
- Enhance validation logic in routes to ensure proper handling of vision models and their requirements.
- Update the AssistantPage and related frontend components to include options for enabling vision understanding and selecting appropriate vision models.
- Modify database seed scripts to include vision-related data for assistants, ensuring consistent setup.
- Refactor related functions to integrate vision model handling in the audio-visual processing pipeline.

backend/db/models.py

@@ -97,6 +97,12 @@ class Assistant(Base):
     runtime_mode: Mapped[str] = mapped_column(String(16), default="pipeline")
     greeting: Mapped[str] = mapped_column(String(2048), default="")
     enable_interrupt: Mapped[bool] = mapped_column(Boolean, default=True)
+    vision_enabled: Mapped[bool] = mapped_column(Boolean, default=False)
+    vision_model_resource_id: Mapped[str | None] = mapped_column(
+        String(40),
+        ForeignKey("model_resources.id", ondelete="SET NULL"),
+        nullable=True,
+    )
 
     # KB 引用:被引用时禁止删 KB(RESTRICT),无默认兜底
     knowledge_base_id: Mapped[str | None] = mapped_column(