fix: entity results in detect_emerging_topics now get related_keywords

Root cause: kw_cooccur was indexed by keyword->co-occurrences. Looking
up an entity name in kw_cooccur only worked if that entity also appeared
as a keyword in some cluster — which most don't. Entity results always
returned empty related_keywords.

Fix: added entity_kw_cooccur dict (entity->Counter[keyword]) populated
in the same per-cluster pass. Entity scoring now uses entity_kw_cooccur
to find top co-occurring keywords. The kw_cooccur dict remains used for
keyword<->keyword co-occurrence in keyword results.

Scope: all detect_emerging_topics results now carry meaningful
related_keywords — calculated from the same timeframe clusters.

news_mcp/mcp_server_fastmcp.py

@@ -711,6 +711,7 @@ async def detect_emerging_topics(limit: int = 10, timeframe: str = "24h", topic:
     kw_sources: dict[str, set] = {}
     kw_buckets: dict[str, set] = {}
     kw_cooccur: dict[str, Counter] = {}
+    entity_kw_cooccur: dict[str, Counter] = {}  # entity -> Counter of co-occurring keywords
 
     bucket_size_hours = max(1.0, hours / 6.0)  # split window into ~6 buckets
 
@@ -807,11 +808,15 @@ async def detect_emerging_topics(limit: int = 10, timeframe: str = "24h", topic:
                     continue
                 kb = kws_in_cluster[j]
                 kw_cooccur[ka][kb] += 1
-            # also track entity↔keyword co-occurrence
+            # also track entity<->keyword co-occurrence (bidirectional)
             for ent in ents_norm:
                 if _is_generic_entity(ent):
                     continue
                 kw_cooccur[ka][ent] += 1
+                # and the reverse: entity -> keyword
+                if ent not in entity_kw_cooccur:
+                    entity_kw_cooccur[ent] = Counter()
+                entity_kw_cooccur[ent][ka] += 1
 
         # bigram phrases (recent only)
         if is_recent:
@@ -866,10 +871,9 @@ async def detect_emerging_topics(limit: int = 10, timeframe: str = "24h", topic:
                     related.append(other)
 
         related_kws = []
-        if ent in kw_cooccur:
-            for kw, _cnt in kw_cooccur[ent].most_common(5):
-                if kw != ent:
-                    related_kws.append(kw)
+        if ent in entity_kw_cooccur:
+            for kw, _cnt in entity_kw_cooccur[ent].most_common(5):
+                related_kws.append(kw)
 
         scored.append({
             "topic": ent,