[EPIC]: Port `std` equivalent parallel algorithms from thrust to `cuda::std`

We want to move thrust parallel algorithms that are equivalent to the standard provided ones to `cuda::std`

This has the clear benefit that we can rely on the familiarity of standard algorithms for C++ users and also simplifies the library design of CCCL.

This issue can be closed by merging implementations of all `cuda::std::` parallel algorithms that are implementable with existing `cub::Device*` algorithms (see list of sub-issues below)

- Can be included via `#include `
- `cuda::execution::gpu` (name TBD) execution policy for targeting GPU backend implemented with CUB algorithms
- This execution policy should be synchronous and need not support passing streams nor memory resources (can be done in future work)
- Use of the other standard execution policies (seq, par, par_unseq, unseq) should fail to compile
- Algorithms should be host-only and do not support CUDA dynamic parallelism
- Algorithms will not support CUDA Graph Capture