Download model gguf file from hf

[phymbert]

Motivation

Since GGUF is officially supported on huggingface <3, '--model' might check for local path then fall back to hf repo id.
If a '--file' is provided it should override default for f16.

References

• llama : auto download HF models if URL provided #4735
• scripts : add hf.sh helper script #5501
• https://huggingface.co/docs/hub/gguf

[phymbert]

@ggerganov thoughts ?

[ggerganov]

We considered this at some point (#4735) but it opens a lot of security concerns due to the core library having to execute commands. So we settled with #5501 which should be good enough for most cases