[v0.9.1][bugfix] disable the chunked prefill feature in Non-MLA LLMs

[rjg-lyh]

What this PR does / why we need it?

This PR enforces the forcible disabling of the chunked prefill feature in Non-MLA models, as the performance of operators supporting this functionality is currently suboptimal.
At the same time, in engine v1 mode, the ascend scheduler is forcibly enabled, and the enable_chunked_prefill specified by the user in additional_config is disabled.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

CI passed with new added/existing test.

[gemini-code-assist]

Code Review

This pull request aims to disable the chunked prefill feature for Non-MLA models due to performance issues. The changes correctly modify the configuration to disable this feature. However, there is a critical bug in the implementation that will cause a NameError when running with Non-MLA models on the v1 engine. I've provided a suggestion to fix this issue.

[gemini-code-assist]

The variable ascend_scheduler_config is not defined at this point, which will lead to a NameError when this code path is executed. It seems you intended to modify the ascend_scheduler_config object within the ascend_config. The correct way to do this would be to access it via ascend_config.ascend_scheduler_config.

Suggested change

	ascend_scheduler_config["enable_chunked_prefill"] = False
	ascend_config.ascend_scheduler_config.enable_chunked_prefill = False

[Yikun]

Current 0.9.1-dev lastest commit: 5926225

Case 1 git revert 9dc23b6, test
Case 2 apply #2659, test (green part latency +0.5 ms)

There are still stable 0.5 ms perf regression compare with your PR and revert #2326

# Start vLLM V1
export MODEL=Qwen/Qwen3-8B
VLLM_USE_V1=1 VLLM_USE_MODELSCOPE=true python3 -m vllm.entrypoints.openai.api_server --model $MODEL \
         --tensor-parallel-size 1 --swap-space 16 --disable-log-stats \
         --disable-log-requests  --load-format dummy

# Benchmark
docker exec -it  yikun-091 bash
export MODEL=Qwen/Qwen3-8B
export VLLM_USE_MODELSCOPE=true
pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
pip install -r /vllm-workspace/vllm-ascend/benchmarks/requirements-bench.txt
python3 /vllm-workspace/vllm/benchmarks/benchmark_serving.py --model $MODEL --dataset-name random \
         --random-input-len 200 --num-prompts 200 --request-rate 1 \
         --save-result --result-dir ./