Handle output from TOSA reference model being int8 (#6190)

Summary:
Move TOSA reference model pin. This adds bugfix for .npy array to now be aligned to output int8 for BI models.

Make sure the dequantization not overflows by adding a cast to int32.

Change-Id: I2274955bc02a6f3a75687bb2f731b2058834c44f

Pull Request resolved: https://github.com/pytorch/executorch/pull/6190

Reviewed By: kirklandsign

Differential Revision: D64426319

Pulled By: digantdesai

fbshipit-source-id: e1169b9420258934ebb82e1148c97ab0f1d206f7