[Bug fix] fix bug found in regression when view chunks for document that not parsed in infinity, it would fail in UI (#14168)

### What problem does this PR solve?
See title, the fail image:
<img width="2667" height="915" alt="20260416-205718"
src="https://github.com/user-attachments/assets/0c564237-5ed0-49af-bf4c-d3b5519abc6e"
/>

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

common/doc_store/infinity_conn_base.py

@@ -491,17 +491,29 @@ class InfinityConnectionBase(DocStoreConnection):
         return len(res)
 
     def get_doc_ids(self, res: tuple[pd.DataFrame, int] | pd.DataFrame) -> list[str]:
+        # Extract DataFrame from result
         if isinstance(res, tuple):
-            res = res[0]
-        return list(res["id"])
+            df, count = res
+            if count == 0:
+                return []
+        else:
+            df = res
+        return list(df["id"])
 
     @abstractmethod
     def get_fields(self, res: tuple[pd.DataFrame, int] | pd.DataFrame, fields: list[str]) -> dict[str, dict]:
         raise NotImplementedError("Not implemented")
 
     def get_highlight(self, res: tuple[pd.DataFrame, int] | pd.DataFrame, keywords: list[str], field_name: str):
+        # Extract DataFrame from result
         if isinstance(res, tuple):
-            res = res[0]
+            df, _ = res
+        else:
+            df = res
+
+        if df.empty or field_name not in df.columns:
+            return {}
+
         ans = {}
         num_rows = len(res)
         column_id = res["id"]