Due to a race condition between rta_apply_hostentry() and rt_update_hostentry(),
happening when a new route is inserted to a table, this commit makes it mandatory
to lock the next hop resolution table while resolving the next hop.
This may be slow, we'll fix it better in some future release