[FEA]: CCCL.c.parallel Should Support Stenciled Transforms

### 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

cuda.cccl.parallel (Python)

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

CUDF uses bitmasks to represent the nullness of rows.
When performing a transform operation, we presently use a pre-computed bitmask computed from the input columns and then run another kernel to compute the output result.
When computing the transform, we skip the input rows that are nulls.

### Describe the solution you'd like

Add a stencil parameter to the `cccl_device_unary_transform_build_ex` and `cccl_device_binary_transform_build_ex` functions.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_