pg_restore --index is as old as e8f69be054e9, so it's not like we could just remove it, but I'd say that with the schema-level restore this would be tempting.
That temptation tells me that there is no appetite for a similar flag for named extended stats objects, which makes me wonder how somebody would get all of the extended stats objects and the pg_restore_extended_stats() calls for a given table, aside from manually filtering the output of a --no-data --statistics dump. It's not a very common use case right now, but I suspect the cases for it will increase in the future.
Anyway, let's improve this situation with the attached, for HEAD and v18.
