[Executorch] mul broadcast update (#6589)

Pull Request resolved: https://github.com/pytorch/executorch/pull/6520

Handle broadcast for > 2D tensors in optimized library. For now broadcast
across only non 0th and (N-1)st dim is supported in optimized path.
ghstack-source-id: 251021628
@exported-using-ghexport

Differential Revision: [D64156862](https://our.internmc.facebook.com/intern/diff/D64156862/)

---------

Co-authored-by: Kimish Patel <kimishpatel@fb.com>