While reviewing Jan's sysdb patches I noticed that a call to
sysdb_attrs_add_uint32() I added recently is wrong.
The reason the code worked and the compiler did not complain was simple
- there's only one place that calls sdap_store_group_with_gid() and it
passes group_attrs for mem_ctx. If we ever reused
sdap_store_group_with_gid() with a different memory context it would
probably crash and burn.
Sorry about the inconvenience.