[FEA]: Vectorize memory operations in CUB 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

CUB

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

Some CUB algorithms (histogram and reduce) rely on vectorized loads. Other CUB algorithms demonstrate poor memory utilization when processing 1-2 byte types.

#### U8


#### U16


### Describe the solution you'd like

We should consider vectorizing memory operations in the following algorithms:

### Tasks
- [ ] Vectorize memory operations in run length encode
- [ ] Vectorize memory operations in partition
- [ ] https://github.com/NVIDIA/cccl/issues/310
- [ ] Vectorize memory operations in scan
- [ ] Vectorize memory operations in reduce by key

### Describe alternatives you've considered

_No response_

### Additional context

_No response_