[FEA]: cuda::barrier:arrival_token should not be constructible from integral types

### 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.

The type of `barrier​::​arrival_token` is unspecified by the standard: http://eel.is/c++draft/thread.barrier#class-7

Currently, constructing an arrival token from primitive types works because it is a type synonym for bool, uint64_t, or __phase_t, depending on the barrier. This code may break in the future.

### Describe the solution you'd like

We should help users to not write non-portable code by by using a wrapper struct instead.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_