server : add cache reuse card link to help

[ggerganov]

Add a link to a visual card in the help for the --cache-reuse argument:

$ llama-server --help

--cache-reuse N                         min chunk size to attempt reusing from the cache via KV shifting
                                        (default: 0)
                                        [(card)](https://ggml.ai/f0.png)
                                        (env: LLAMA_ARG_CACHE_REUSE)

This link will open the following image:

Could be useful to click on it from the terminal to get some extra information about how the feature works. Maybe we can improve the help by adding more cards like this.

[ericcurtin]

Cool, I've learned what cache reuse is now 😄

[ngxson]

I'm a bit worry that the image will take a large amount of space on rendering to the server/README file. I think leaving just a link is fine.

Also, could we host the image (or do a proxy) via ggml.ai domain to make the URL looks shorter?

[ggerganov]

I'm a bit worry that the image will take a large amount of space on rendering to the server/README file. I think leaving just a link is fine.

The readme only has the link - it does not render the image currently. I agree rendering the images would be too heavy.

Also, could we host the image (or do a proxy) via ggml.ai domain to make the URL looks shorter?

Good idea: https://ggml.ai/f0.png