fix: Parse actual image URLs from search HTML for thumbnails

Extract main_image from HTML img tags instead of constructing from
car_no pattern. Handles timestamp-based filenames (e.g., 1767925381_0.jpg)
that differ from the default cmcar_0.jpg pattern.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/app/api/carmodoo.py

@@ -516,7 +516,21 @@ class CarmodooClient:
                         except:
                             pass
 
-                    # 이미지 URL 생성
+                    # 이미지 URL: HTML img 태그에서 실제 URL 추출 (타임스탬프 패턴 대응)
+                    main_image = ""
+                    img_elems = row.xpath('.//img[contains(@src, "__carPhoto")]/@src')
+                    if img_elems:
+                        img_src = img_elems[0]
+                        # __THUM 접미사 제거하여 원본 이미지 URL 생성
+                        img_src = img_src.replace('__THUM', '')
+                        if img_src.startswith('/'):
+                            main_image = f"{CARMODOO_BASE_URL}{img_src}"
+                        elif img_src.startswith('http'):
+                            main_image = img_src
+                        else:
+                            main_image = f"{CARMODOO_BASE_URL}/{img_src}"
+                    if not main_image:
+                        # 폴백: 기존 패턴
                         car_no_padded = car_no.zfill(9)
                         main_image = f"{CARMODOO_BASE_URL}/data/__carPhoto/{car_no_padded[:3]}/{car_no_padded[3:6]}/{car_no_padded[6:]}/cmcar_0.jpg"