[FEA]: Update standalone header testing to use installed version of CCCL

### 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

Infrastructure

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

As a user and maintainer of CCCL, I want to be sure that all public headers are usable in an installed version of CCCL.

Today, CCCL tests all public headers can be `#include ` into an empty TU and compiled successfully.

However, this testing is done in the context of the GitHub repo, *not* the installed versions of the headers. This can cause problems if our install rules are misconfigured and forget to install some of the headers.

### Describe the solution you'd like

We should update the per-header include testing for Thrust/CUB/libcu++ to work off an _installed_ version of CCCL.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

Start by locating the existing per-header include tests for Thrust, CUB, and libcu++, then trace the installation rules that produce CCCL's installed headers. Update the test flow so it compiles against the installed version rather than the repository checkout. Done means every public header include test passes from the installed CCCL tree.

Written by the indexing model from the issue text.