Update moe_align kernel

[charlifu]

Update moe_algin kernel to the latest vllm version.

This PR is for #107

[yiakwy-xpu-ml-framework-team]

@charlifu Could you kindly have a look this pending PR ? It is AMD friendly kernel, very fast.

As for copyright issue , note the original kernel from vllm contributed from SGLang authors according to MIT liencse for compilance requirements.

As far as I understand (fix me if I am wrong), SGlang team will use more and more utilities from Flashinfer. For example vec_t for different dtype vectorization (uint4 128 bit load).

The problems is vec_t does not handle unligned cases (the original design is leave to user to handle it) :

offset = load<srcDtype ,4>(src, dst, /*offset*/0);
// handling 
offset = load<srcDtype, 2>(src, dst, offset);
offset = load<srcDtyep, 1>(src, dst, offset);

Luckily there is no need to dispatch fp8/fp16/bf16 for this kind of kernel. But later we still need this capability.

Consider the status of flashinfer-rocm, I prefer to use CK vectorization capability to do this later.