[FEA] Support spill+retry for GpuExpandExec

[revans2]

Is your feature request related to a problem? Please describe.
GpuExpandExec sadly keeps a ColumnarBatch around between calls. Not a SpillableGpuColumnarBatch or anything like that.

We should update it to use a SpillableColumnarBatch. Support tiered project so that we can get the retry easily.

We also have a lot of code in there for handling columnaEval returning a null, but it should never return a null it would return a GpuScalar, so we should rip out that code. It was a premature optimization that also could cache data in memory that is not spillable. So two wrongs are making a right here.

This really only runs right before a hash aggregate, but with the retry logic in hash aggregate it is kind of scary and I think we should fix this sooner than later.

[revans2]

This was fixed by #8467