[FEA]: Implement GPU friendly single-threaded sorting algorithms

### 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

libcu++

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

We currently do not expose `cuda::std::` sorting algorithms, because they are commonly implemented through recursion and that might blow up the GPU stack.

We need to spend time to implement a proper GPU friendly, single-threaded implementation to expose
* `sort`
* `stable_sort`
* `nth_element`
* `inplace_merge`

### Describe the solution you'd like

There are a lot of nice sorting networks available, that are already in the libc++ implementation.

We should investigate if we can use some larger ones to properly implement a non-recursive sort on device

Note that we do not want to have implementation divergence between host and device

Also this is regarding *serial* algorithms. For parallel algorithms we will use the CUB backend

### Describe alternatives you've considered

_No response_

### Additional context

_No response_