plugin JID Search: Update cutoff ratio + dedup:

- `RATIO_CUTOFF` has been adapted to avoid returning approximate matches.
- In case of duplicate entries (e.g., rooms returned by the local search are also returned
  by the external one), we keep only the item with the most metadata.
- Exclude `None` values in `jid_search` bridge method.

libervia/backend/plugins/plugin_misc_jid_search.py

@@ -45,7 +45,8 @@ PLUGIN_INFO = {
     C.PI_HANDLER: "no",
     C.PI_DESCRIPTION: _("""Search for XMPP entities"""),
 }
-RATIO_CUTOFF = 0.6
+# Minimal ratio for fuzzy matching
+RATIO_CUTOFF = 0.75
 # Used when a search term matches a substring of a relevant data.
 PARTIAL_MATCH_RATIO = 0.8
 MAX_CACHE_SIZE = 10
@@ -99,6 +100,24 @@ class SearchItems(RootModel):
         self.root.sort(key=key, reverse=reverse)  # type: ignore
 
 
+def _data_rank(item: SearchItem) -> tuple[bool, bool, bool, bool]:
+    """Rank a search item by its metadata completeness
+
+    We can have duplicate items when allowing external search (local MUC can be referenced
+    globally). To choose the item to keep, we check the one with most metadata, and chose
+    the local one in case of equivalent data.
+
+    @param item: The item to rank.
+    @return: A tuple to use for comparaison.
+    """
+    return (
+        getattr(item, "nusers", None) is not None,
+        bool(getattr(item, "description", None)),
+        bool(getattr(item, "language", None)),
+        getattr(item, "local", False),
+    )
+
+
 class Options(BaseModel):
     entities: bool = Field(
         default=False, description="Search for entities for direct chat"
@@ -137,7 +156,9 @@ class JidSearch:
         client = self.host.get_client(profile)
         options = Options.model_validate_json(options_s)
         d = defer.ensureDeferred(self.search(client, search_term, options))
-        d.addCallback(lambda search_items: search_items.model_dump_json())
+        d.addCallback(
+            lambda search_items: search_items.model_dump_json(exclude_none=True)
+        )
         return d
 
     async def search(
@@ -217,6 +238,20 @@ class JidSearch:
                     )
                 )
 
+        # Deduplicate by entity, keeping the most informative item (the global directory
+        # often provides more data)
+        by_entity = {}
+        for item in matches:
+            existing = by_entity.get(item.entity)
+            if existing is None:
+                by_entity[item.entity] = item
+            elif _data_rank(item) > _data_rank(existing):
+                # the entity keeps the best match score of its duplicates
+                item.exact_match = item.exact_match or existing.exact_match
+                item.relevance = max(item.relevance or 0, existing.relevance or 0) or None
+                by_entity[item.entity] = item
+        matches = SearchItems(list(by_entity.values()))
+
         matches.sort(
             key=lambda item: (
                 item.exact_match,