url: save -mu downloads to new cache location

[ochafik]

This makes -mu (updated in #6930) save/lookup files in the cache just like -hfr + -hff (#7353) by default (unless -m is provided explicitly). Also see #7252

./server -mu https://huggingface.co/NousResearch/Meta-Llama-3-8B-Instruct-GGUF/resolve/main/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf

After this gets merged, if you have already downloaded .gguf / .gguf.json file pairs to models/, simply move them to the cache folder:

• $LLAMA_CACHE if defined,
• ~/Library/Caches/llama.cpp/ on Mac,
• ~/.cache/llama.cpp on Linux (or $XDG_CACHE_HOME/llama.cpp if defined)
• %LOCALAPPDATA%\llama.cpp on Windows

[ngxson]

LGTM, only a small suggestion regarding naming.