[AMD] Optimize search space and upgrade Image to 0.19.0 for MiniMax-M2.5

[chunfangamd]

• Optimize MiniMax-M2.5 FP8 MI355X vLLM search-space
• Upgrade Image to v0.19.0
• Enable FP8 KV cache + AITER FA (It is important to recheck this when upgrading to higher image versions)

e2e Test run: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/23987768210

co-author: @benenzhu

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://cookbook.sglang.io/
• https://docs.vllm.ai/projects/recipes/en/latest/

If it is not, please create a PR first before we can merge your PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sgl-cookbook

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://cookbook.sglang.io/
• https://docs.vllm.ai/projects/recipes/en/latest/

If it is not, please create a PR first before we can merge your PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sgl-cookbook

[functionstackx]

lgtm once validation passes

[benenzhu]

vllm recipes pr updated: vllm-project/recipes#300