Propagate mul optimizations from D61504544/D61560825/D61560826 to add/sub/div (#4966)

These diffs only updated the optimized mul operator; make the corresponding changes to the other binary arithmetic operators.

Differential Revision: D61577411

Pull Request resolved: https://github.com/pytorch/executorch/pull/4816

---------

Co-authored-by: Scott Wolchok <swolchok@fb.com>