[FEA]: Higher precision calculations for validation purposes

### 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

Not sure

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

With the `awkward` library (https://github.com/scikit-hep/awkward), we are attempting to validate operations with the `cuda` backend against the same operations with the `cpu` backend.

@ianna and I have noticed that in some cases, there appear to be differences in precision between CPU and GPU, as documented here: https://github.com/scikit-hep/awkward/issues/3711

These differences can make validation difficult.

### Describe the solution you'd like

We would like an option to retain more precision in the GPU computations (even if this means the computation is slower), to facilitate validation against CPU operations.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_