[BUG]: Thrust execution policy allocators should be rebound to the target type before allocating.

### Is this a duplicate?

- [X] I confirmed there appear to be no [duplicate issues](https://github.com/NVIDIA/cccl/issues) for this bug and that I agree to the [Code of Conduct](CODE_OF_CONDUCT.md)

### Type of Bug

Runtime Error

### Component

Thrust

### Describe the bug

Thrust currently uses execution policy allocator as-is, without rebinding ([example](https://github.com/NVIDIA/cccl/blob/8cd246458ffed4894632c7e7ac566686288174f9/thrust/thrust/detail/execute_with_allocator.h#L53-L63)). This may result in type-specific issues, such as alignment problems.

Ideally, an execution policy should be able to be created with a `void`-typed allocator without issue, but currently this will fail to compile as the code linked above attempts to calculate `sizeof(void)`.

### How to Reproduce

Execute `thrust::reduce(thrust::device(thrust::device_allocator{}), ...)`. This will fail to compile.

### Expected behavior

Passing a `void`-typed allocator should compile, and all allocations using this generic allocator should be rebound prior to use.

### Reproduction link

_No response_

### Operating System

_No response_

### nvidia-smi output

_No response_

### NVCC version

_No response_

Start with thrust/thrust/detail/execute_with_allocator.h at the linked allocation code, then reproduce the compile failure with thrust::reduce(thrust::device(thrust::device_allocator{}), ...). Done means a void-typed execution-policy allocator compiles and generic allocations are rebound to their target type before use.

Written by the indexing model from the issue text.