Why is there no way to allocate local memory inside a ND-kernel (parallel_for)?

I know this is possible using the hierarchical `parallel_for` invoke, but we can't do it with the `nd_item` version despite being able to specify the group size. OpenCL allows this (AFAIK), SYCL's competitors do, too, so why not allow it in SYCL? I suppose there must be a reason for leaving it out.

Start with the nd_item and hierarchical parallel_for forms referenced in the issue, then compare their local-memory capabilities with OpenCL. A complete result would establish and document a specification decision about local memory in the nd_item version; no file or test is identified in the payload.

Written by the indexing model from the issue text.