KV Cache simulation for calculation of request prefill time

[irar2]

For a request, check if we already have a part of it in the local kv cache, and only use the remaining number of tokens in prefill time calculation