[inductor] Lower aten.prod (#99484)
This lowers `aten.prod` using the new `tl.reduce` functionality in triton. I
also introduce `TritonKernel.helper_functions` which allows code to be defined
outside of the kernel body so that we can defined the `_prod_accumulate` helper
function.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99484
Approved by: https://github.com/ngimel
