[FEA]: Add tests that verify Thrust/CUB algorithms satisfy necessary requirements about how many times they access inputs/outputs

### 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

General CCCL

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

As a user of Thrust/CUB algorithms, I want to be sure that they satisfy any explicit/implicit requirements about how many times elements in the input/output range are dereferenced.

For example, `for_each` must guarantee that each input is only dereferenced exactly once to allow the operator to have side-effects. We don't actually test that today.

### Describe the solution you'd like

We should add testing utilities that verify input/output iterators are dereferenced only once and use those in the appropriate algorithms (e.g., `for_each`).

### Describe alternatives you've considered

_No response_

### Additional context

Today, `for_each` is the only algorithm where we fundamentally require the input elements are only dereferenced once. However, there is interest in potentially expanding this to other algorithms as needed or as possible.