[FEA]: Analyze segment_sort/keys benchmark performance

### Is this a duplicate?

- [x] I confirmed there appear to be no [duplicate issues](https://github.com/NVIDIA/cccl/issues) for this request and that I agree to the [Code of Conduct](CODE_OF_CONDUCT.md)

### Area

cuda.compute (Python)

### Is your feature request related to a problem? Please describe.

Follow up from the new nvbench Python benchmarks and comparison with C++ ones: https://github.com/NVIDIA/cccl/pull/7341

We are getting about 90% performance of C++ on the tranform/heavy benchmark.

### Describe the solution you'd like

We have a target of 5% performance gap with C++ for large amount of items

### Describe alternatives you've considered

_No response_

### Additional context

_No response_