[Arm] Support Linear INT16 TOSA reference model run

[Ninja91]

🚀 The feature, motivation and pitch

With this PR: https://github.com/pytorch/executorch/pull/13658/files we add support and test to 16a8w_quantize and lower Linear op. This fails at tosa reference model run and been marked as xfail. Enable run on TOSA ref model and enable this test.
cc: @gggekov , @digantdesai , @3l1

Alternatives

No response

Additional context

#13635

RFC (Optional)

No response

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218