use selected index past past key value in attention when using contin…

[Eayne]

when using continuous kv cache, gpt_attention will only use first past_key_value instead of past_key_value[selected_indexed]. It will cause calculating result errors when the values of continous kv caches are not zeros.

[juney-nvidia]

@Eayne

Hi, since TensorRT-LLM becomes github firstly since last Monday, pls refresh your MR based on the latest main if you still want to contribute this.

Thanks June

[poweiw]

Closing since no response after https://github.com/NVIDIA/TensorRT-LLM/pull/2682#issuecomment-2761036095. Feel free to reopen! @Eayne