update readme with the latest hf transformers

…ng new kernels (#169)

This PR does not change any logic or functionality in the main branch.
It restructures the existing C++ code organization to facilitate the
addition of new kernels in subsequent PRs.

In general, each customized CUDA kernel involves modifying the following
three files at the backend:

1. Registering Python bindings and adding declarations in `ops.cc`.
1. Placing the host's kernel launch function in `x.cc`, where `x` is the
operator's name.
1. Implementing CUDA kernels in `x.cuh`.

check res_index before print it.

add
[aphrodite-engine](https://github.com/aphrodite-engine/aphrodite-engine)
in news

add index bitwidth=10