Small speed up joining large remote rooms (#9825)

There are a couple of points in `persist_events` where we are doing a
query per event in series, which we can replace.

changelog.d/9825.misc

@@ -0,0 +1 @@
+Small speed up for joining large remote rooms.

synapse/storage/databases/main/events.py

@@ -1378,16 +1378,20 @@ class PersistEventsStore:
             ],
         )
 
-        for event, _ in events_and_contexts:
-            if not event.internal_metadata.is_redacted():
         # If we're persisting an unredacted event we go and ensure
         # that we mark any redactions that reference this event as
         # requiring censoring.
-                self.db_pool.simple_update_txn(
+        sql = "UPDATE redactions SET have_censored = ? WHERE redacts = ?"
-                    txn,
+        txn.execute_batch(
-                    table="redactions",
+            sql,
-                    keyvalues={"redacts": event.event_id},
+            (
-                    updatevalues={"have_censored": False},
+                (
+                    False,
+                    event.event_id,
+                )
+                for event, _ in events_and_contexts
+                if not event.internal_metadata.is_redacted()
+            ),
         )
 
         state_events_and_contexts = [
@@ -1881,18 +1885,26 @@ class PersistEventsStore:
                 ),
             )
 
-        for event, _ in events_and_contexts:
+            room_to_event_ids = {}  # type: Dict[str, List[str]]
-            user_ids = self.db_pool.simple_select_onecol_txn(
+            for e, _ in events_and_contexts:
+                room_to_event_ids.setdefault(e.room_id, []).append(e.event_id)
+
+            for room_id, event_ids in room_to_event_ids.items():
+                rows = self.db_pool.simple_select_many_txn(
                     txn,
                     table="event_push_actions_staging",
-                keyvalues={"event_id": event.event_id},
+                    column="event_id",
-                retcol="user_id",
+                    iterable=event_ids,
+                    keyvalues={},
+                    retcols=("user_id",),
                 )
 
-            for uid in user_ids:
+                user_ids = {row["user_id"] for row in rows}
+
+                for user_id in user_ids:
                     txn.call_after(
                         self.store.get_unread_event_push_actions_by_room_for_user.invalidate_many,
-                    (event.room_id, uid),
+                        (room_id, user_id),
                     )
 
         # Now we delete the staging area for *all* events that were being