Support async in DeepEP

[fzyzcjy]

Motivation

When doing #4068, DeepEP needs to be async. This PR enables that in a minimal way.

(This is a separate PR because #4068 may not be done in a day, and I hope less merge conflicts happen, so extract this part first)

Related: #4232 (Initial DeepEP support)

When viewing diff, please subtract from change in #4608

In order to demonstrate the PR works, I set async_finish=True temporarily. In real world, maybe async will be true only when doing two-batch-overlap. The flag will be handled in #4068. (I can also make it false in this PR if needed)

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[ch-wan]

For future reference, I did a simple benchmark on one 8xH200 machine for this PR. Here is the command

python3 -m sglang.launch_server --model-path deepseek-ai/DeepSeek-V3 --trust-remote-code   --tp 8 --dp 8 --host 0.0.0.0 --port 30000   --enable-dp-attention --enable-deepep-moe   --disable-cuda-graph
python3 -m sglang.bench_serving --backend sglang --dataset-name random --num-prompt 512 --random-input 1000 --random-output 1000 --random-range-ratio 1 --host 127.0.0.1 --port 30000 --max-concurrency 128

Results:

Version	Concurrency	Input	Output	Num Requests	Input Throughput(tok/s)	Output Throughput (tok/s)	Total Throughput (tok/s)
Oritinal	127.98	1000	1000	512	555.48	555.48	1110.97
Current	127.96	1000	1000	512	612.16	612.16	1224.33