### What this PR does / why we need it?
Implement `apply_top_k_top_p` via ascendC to eliminate the constraint of
k [1,1024]. It enables high performance TopKTopP calculation and avoid
D2H synchronization introduced by k validation.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
E2E serving with `k=4096` and `p=0.95`
- vLLM version: v0.13.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/d68209402ddab3f54a09bc1f4de9a9495a283b60
---------
Signed-off-by: linfeng-yuan <1102311262@qq.com>
Signed-off-by: SlightwindSec <slightwindsec@gmail.com>
Co-authored-by: SlightwindSec <slightwindsec@gmail.com>