[FEA]: CCCL.c.parallel should support n-ary 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.

CCCL.c.parallel's transform only supports unary and binary transforms.
In [CUDF](https://github.com/rapidsai/CUDF), a single column could have 3 input arrays, an example being the string column, which is composed of prefix-sum indices, string data, and null-mask.
Most data transformation operations involve more than one column and are of arbitrary counts.

This is important for CUDF's adoption

### Describe the solution you'd like

Implement multi-input & multi-output transforms on an arbitrary number of iterators (preferably unbounded).

### Describe alternatives you've considered

None will be feasible; the iterators are type and data-rate specific.

### Additional context

CUDF's Multi-input Transform support PR: https://github.com/rapidsai/cudf/pull/17881
Blogpost with some more examples of multi-input transform use-cases: https://developer.nvidia.com/blog/efficient-transforms-in-cudf-using-jit-compilation/