[FEA] Improve GDS spill performance on NVIDIA EGX

[rongou]

Is your feature request related to a problem? Please describe.
GDS spill does not perform as well on EGX clusters as compared to DGX. There are a couple of potential causes tied to hardware configuration:

• EGX machines have more RAM (host memory) per GPU, so for a lot of queries spilled buffers can be mostly held in host memory, with only a small portion spilled all the way to disk. Currently GDS spill always spills everything to disk, so it's at a disadvantage.
• EGX machines usually have fewer NVMe drives compared to DGX, so IO bandwidth is smaller.

Describe the solution you'd like
One possibility is to combine GDS spill with host memory. Device buffers can be spilled to host memory first, and when the host memory spill storage limit is reached, then spill through GDS to disk.

Describe alternatives you've considered
We can try to continue improve GDS spill performance, but in a head to head comparison GDS is probably never going to be faster than host memory.

Additional context
Current spilling logic is mostly in RapidsHostMemoryStore and RapidsGdsStore.