feat: chat uploads, vision turn isolation, and MindSpace agent improvements

Add chat file/image upload UX, attachment proxying, vision thumbnails, and per-turn image scoping so agents only use the current upload. Extend MindSpace asset context, billing token state, OA/scenario verify scripts, and related runtime config.

Co-authored-by: Cursor <cursoragent@cursor.com>

llm-providers.mjs

@@ -2025,7 +2025,9 @@ export function createLlmProviderService(
       const content = [
         ...imageItems.map((item) => ({
           type: 'image_url',
-          image_url: { url: `data:${item.mimeType};base64,${item.data}` },
+          image_url: {
+            url: `data:${item.visionMimeType ?? item.mimeType};base64,${item.data}`,
+          },
         })),
         { type: 'text', text: String(userText || '请描述这张图片的内容').trim() },
       ];