On The User-Scheduler Relationship In High-Performance Computing