Minor receive cleanup, queue empty buffers to the hardware in 
groups of 16 to reduce unneeded fetches and improve PCI 
efficiency.