Added vectorized flip for uint8 (#90013)
Following https://github.com/pytorch/pytorch/pull/89414#discussion_r1036224613 just refactoring and adding `flip` method for `Vectorized<uint8>`. This should speed up torch.flip horizontal implementation similarly to what is reported in https://github.com/pytorch/pytorch/pull/89414
Pull Request resolved: https://github.com/pytorch/pytorch/pull/90013
Approved by: https://github.com/peterbell10, https://github.com/lezcano
