[DOC]: Add how-to guide for launching CUB algorithms with CUDA Graphs

### Is this a duplicate?

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

### Is this for new documentation, or an update to existing docs?

New

### Describe the incorrect/future/missing documentation

CUB users may want to integrate CUB device-wide algorithms into CUDA Graph workflows, but the documentation does not currently provide a focused guide showing the recommended launch pattern.

We should add a CUB documentation page that demonstrates how to launch a CUB algorithm as part of a CUDA Graph, including temporary storage setup, stream capture, graph instantiation, graph launch, and result validation.

The guide should make clear which parts of the workflow happen before capture, which operations are captured, and how users should structure allocation and temporary storage to avoid common graph-capture pitfalls.

### If this is a correction, please provide a link to the incorrect documentation. If this is a new documentation request, please link to where you have looked.

_No response_