Shared-prefix rope issue

[lkc1997]

I found that during shared-prefix calculation, this kenerl won't use qo_indptr to split batch queries which may cause rope error.

[yzh119]

Hi @lkc1997 , we have another few arguments (rope_position to specify the rope position of each query) in our C++ APIs but we have ported them into PyTorch APIs, I'll do that in next release.

You can process rope outside attention and use attention kernel without rope at the moment.