fix `upsample_nearest` decompositions for `uint8` tensors (#106675)
Fixes #106674.
This PR aligns the implementation of `_compute_upsample_nearest_indices` with `UpSampleKernel.cpp`: https://github.com/pytorch/pytorch/blob/68cb854d73458a14684d584c25c22b17eb79dfca/aten/src/ATen/native/cpu/UpSampleKernel.cpp#L1388-L1393
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106675
Approved by: https://github.com/albanD