Add unit test for scalar vs. CGSize1 probing

[sleeepyjack]

This PR adds a unit test to verify that the probing sequences for the scalar probing iterator vs. the iterator that uses a cooperative group of size 1 are actually equivalent.

[sleeepyjack]

Found the culprit with double hashing:
Apparently this expression in the scalar code path is different from this one in the CG code path. I'd have to wrap my head around that logic again to fix this. Alternatively we could internally dispatch to the CG code path in the CG probing iterator when CGSize==1 and call it a day.

[PointKernel]

With #501 fixing the sanitize hash logic, removing max from the scalar probing should fix the issue.

[sleeepyjack]

that fix broke something else: https://github.com/NVIDIA/cuCollections/actions/runs/9584556889/job/26440581801?pr=509#step:7:758

[PointKernel]

that fix broke something else: https://github.com/NVIDIA/cuCollections/actions/runs/9584556889/job/26440581801?pr=509#step:7:758

Which fix are you referring to? removing max from double hashing?

[sleeepyjack]

Which fix are you referring to? removing max from double hashing?

Operator presedence: f0ae693

Not sure where my logic is flawed. 😢

[PointKernel]

Which fix are you referring to? removing max from double hashing?

Operator presedence: f0ae693

Not sure where my logic is flawed. 😢

No good way but printing out all probing steps. I will take a look tomorrow.

[PointKernel]

Should be ready to go.

[sleeepyjack]

You sir are the MVP