[FEA]: Expose non-stable algorithms variants in CUB

### 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

CUB

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

In some cases, users might not care about stability. For instance, if a user is interested in finding elements satisfying a predicate, they'd use select that guarantees the relative order of elements is preserved. If this guarantee is not needed, we might exploit striped loads avoiding memory transposition and leading to better performance.

### Describe the solution you'd like

We should experiment with striped loading in partition, select, and radix sort, and if there are noticeable performance benefits, expose non-stable versions.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_