[BUG] Investigate GpuRand usage outside of ProjectExec

[revans2]

Describe the bug
#11646 was filed for an issue where GpuRand didn't work correctly in all cases. #11647 fixed the issue, but it exposed a problem where we are not doing a checkpoint restore retry in all locations that GpuRand could run. That means if we have a GpuRand outside of a regular project exec we might produce incorrect numbers on a retry.

#11647 (review)

So we need to do a few things for a complete solution

1. We need to go through the Spark code and figure out what are all of the places that a non-deterministic expression could be run. We can do this by looking at all of the places that initialize is called on non-deterministic expressions.
2. We need code changes so that if a retry happens on a non-deterministic expression that is outside of a checkpoint/restore, then we fail instead of retrying.
3. We also want a way to detect a non-deterministic expression being run outside of a checkpoint/restore retry block and throw an error from the plan so that when we can have tests validate that we have this covered.
4. We need a lot more tests to verify that we are doing the right thing with GpuRand.