[Feature]: enable prefix caching when MLA is enabled

[ZhongYingMatrix]

🚀 The feature, motivation and pitch

I notice in #12642, we disable prefix caching when model is with mla (deepseek). Is there any problem of mla coming with prefix caching? If we can enable it by any chance?

Alternatives

No response

Additional context

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[ZhongYingMatrix]

I also notice it's removed in https://github.com/vllm-project/vllm/pull/12639/files#diff-7eaad0b7dee0626bf29d10081b0f0c5e3ea15a4af97e7b182a4e0d35f8346953.
Hi @LucasWilkinson by any chance u got some related information? Really appreciate that~

[ZhongYingMatrix]

fix in #14255