Rename packed tensor accessor (#25654)
Summary:
Closes https://github.com/pytorch/pytorch/issues/19268
This does the renaming suggested by ezyang in https://github.com/pytorch/pytorch/issues/19268#issuecomment-490478887 except that the templated version of `packed_accessor` is also renamed to `generic_packed_accessor`.
Additionally, all of the users I could find in `ATen/native/cuda` are updated without changing their index types.
The corresponding tutorial update is in pytorch/tutorials#644
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25654
Differential Revision: D17259208
Pulled By: ezyang
fbshipit-source-id: 172a46f623d544ca16f7ed5077b6e4f57a3d1f21
