feat: refactor PDF extraction to caption-based screenshots, add upvote refresh, clean up UI

- PDF extractor: rewrite from embedded bitmap extraction to caption-based
  page region screenshots. Finds Figure/Table captions via regex,截取上方/下方
  page region, handles compound figures and vector graphics.
- Upvote refresh: new crawler.refresh_upvotes() re-fetches upvotes for recent
  N days without inserting new papers. Scheduler runs daily 30min after pipeline.
- Admin: add /admin/refresh-upvotes endpoint and dashboard button.
- UI: remove date quick nav, show upvote update time on detail/card pages,
  clean up CSS date-chip styles.
- Utils: add recent_date_strs() helper.

app/services/crawler.py

@@ -16,7 +16,7 @@ from app.models import (
     SummaryState,
     SummaryStatus,
 )
-from app.utils import make_http_client, utc_now
+from app.utils import make_http_client, recent_date_strs, utc_now
 
 logger = logging.getLogger(__name__)
 
@@ -205,3 +205,81 @@ async def crawl_daily(db: Session, target_date: str, top_n: int | None = None) -
         log_entry.completed_at = utc_now()
         db.commit()
         return {"found": 0, "new": 0, "status": "failed", "error": str(exc)}
+
+
+async def refresh_upvotes(db: Session, days: int | None = None) -> dict:
+    """重新抓取最近 N 天论文的 upvotes，不插入新论文。
+
+    遍历每天调用 HF API，对已有论文仅更新 upvotes 和 crawled_at。
+    """
+    days = days or settings.UPVOTE_REFRESH_DAYS
+    date_strs = recent_date_strs(days)
+    now = utc_now()
+
+    log_entry = CrawlLog(
+        task="upvote_refresh",
+        status="running",
+        date=date_type.fromisoformat(date_strs[0]),
+        started_at=now,
+    )
+    db.add(log_entry)
+    db.commit()
+
+    total_updated = 0
+    errors: list[str] = []
+
+    try:
+        for ds in date_strs:
+            try:
+                raw_papers = await fetch_daily(ds)
+                updated = _update_upvotes_only(db, raw_papers)
+                total_updated += updated
+                logger.info("Refreshed upvotes for %s: %d papers", ds, updated)
+            except Exception as exc:
+                msg = f"{ds}: {exc}"
+                errors.append(msg)
+                logger.warning("Failed to refresh upvotes for %s: %s", ds, exc)
+
+        log_entry.status = "success" if not errors else "partial"
+        log_entry.papers_found = total_updated
+        log_entry.details_json = f'{{"days": {days}, "errors": {len(errors)}}}'
+        log_entry.completed_at = utc_now()
+        db.commit()
+
+        return {
+            "status": "success" if not errors else "partial",
+            "updated": total_updated,
+            "days": days,
+            "errors": errors or None,
+        }
+    except Exception as exc:
+        logger.exception("Upvote refresh failed")
+        log_entry.status = "failed"
+        log_entry.error = str(exc)
+        log_entry.completed_at = utc_now()
+        db.commit()
+        return {"status": "failed", "updated": total_updated, "error": str(exc)}
+
+
+def _update_upvotes_only(db: Session, papers_raw: list[dict]) -> int:
+    """对已有论文更新 upvotes，跳过不存在的新论文。"""
+    now = datetime.now(timezone.utc)
+    updated = 0
+
+    for item in papers_raw:
+        meta = _parse_paper(item)
+        arxiv_id = meta["arxiv_id"]
+        if not arxiv_id:
+            continue
+
+        existing = db.execute(
+            select(Paper).where(Paper.arxiv_id == arxiv_id)
+        ).scalar_one_or_none()
+
+        if existing:
+            existing.upvotes = meta["upvotes"]
+            existing.crawled_at = now
+            updated += 1
+
+    db.commit()
+    return updated