Side-by-side image/text layout and editable metadata on review

New Job page:
- OCR result now shows source image and editable textarea side by side
- Grounding-box overlay preview moved into the non-commit branch

Browse Jobs / Review page:
- JobDetail uses a 2-column layout: image + read-only info on left,
  all editable fields on right
- Author, book, chapter, and page are now editable inputs (not read-only)
- Text textarea is always editable (for both unreviewed and reviewed jobs)
- Reviewer name pre-filled for reviewed jobs; button becomes "Save Changes"
- Outer grid changed to 1/3 list + 2/3 detail for more review space

Backend:
- PUT /api/jobs/{id}/review now accepts and saves author, book,
  chapter, page alongside reviewed_text and reviewer_name

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/main.py

@@ -603,6 +603,10 @@ async def process_pdf(
 class ReviewRequest(BaseModel):
     reviewed_text: str
     reviewer_name: str
+    author: Optional[str] = None
+    book: Optional[str] = None
+    chapter: Optional[str] = None
+    page: Optional[str] = None
 
 
 def _job_row_to_dict(row) -> Dict[str, Any]:
@@ -811,11 +815,23 @@ async def review_job(job_id: str, body: ReviewRequest):
                     SET status = 'reviewed',
                         reviewed_text = %s,
                         reviewer_name = %s,
-                        reviewed_at = NOW()
+                        reviewed_at = NOW(),
+                        author = %s,
+                        book = %s,
+                        chapter = %s,
+                        page = %s
                     WHERE id = %s
                     RETURNING *
                     """,
-                    (body.reviewed_text, body.reviewer_name, job_id),
+                    (
+                        body.reviewed_text,
+                        body.reviewer_name,
+                        body.author or None,
+                        body.book or None,
+                        body.chapter or None,
+                        body.page or None,
+                        job_id,
+                    ),
                 )
                 row = cur.fetchone()
     except Exception as exc: