[FEA]: Optimize `cub::DeviceUniqueByKey` by conditionally loading values

### 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.

Similar to https://github.com/NVIDIA/cccl/issues/326, CUB UniqueByKey unconditionally loads all values before figuring out which keys are are actually needed.

### Describe the solution you'd like

If all values are not set, we should skip them. An early implementation where I only load after we figure out which values to keep lead to regressions in a few cases, notably for smaller input sizes.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_