Remove use of eval() from search.py (#4887)

Use `json.loads()` instead.

### What problem does this PR solve?

Using `eval()` can lead to code injections. I think this loads a JSON
field, right? If yes, why is this done via `eval()` and not
`json.loads()`?

### Type of change

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

rag/nlp/search.py

@@ -15,6 +15,7 @@
 #
 import logging
 import re
+import json
 from dataclasses import dataclass
 
 from rag.settings import TAG_FLD, PAGERANK_FLD
@@ -258,7 +259,7 @@ def _rank_feature_scores(self, query_rfea, search_res):
         q_denor = np.sqrt(np.sum([s*s for t,s in query_rfea.items() if t != PAGERANK_FLD]))
         for i in search_res.ids:
             nor, denor = 0, 0
-            for t, sc in eval(search_res.field[i].get(TAG_FLD, "{}")).items():
+            for t, sc in json.loads(search_res.field[i].get(TAG_FLD, "{}")).items():
                 if t in query_rfea:
                     nor += query_rfea[t] * sc
                 denor += sc * sc