Introduce uni-tensor and bi-tensor equivalents (#6006)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/6006

This introduces unitensor and bitensor equivalents to the tritensor template introduced by swolchok.
They are applied to add.Scalar_out and add.out respectively, reducing op_add's build size.
The unitensor template is applied to clamp.Scalar_out, reducing op_clamp's build size further.

Build size reduction:
- add: 484K -> 21K
- clamp: 119K -> 28K
ghstack-source-id: 246919722
exported-using-ghexport

Reviewed By: swolchok

Differential Revision: D63838076

fbshipit-source-id: 564d4435e51302856611b309ac0ef957baee3a43