Fix HF_HUB_OFFLINE=1 for Gaudi backend

[regisss]

What does this PR do?

Fixes #3192.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[regisss]

@Narsil I added a small change to router/src/server.rs. I realized the cache path was sometimes not set properly.

It appears that, when we have HF_HUB_OFFLINE=1, the cache is set to HUGGINGFACE_HUB_CACHE:

text-generation-inference/router/src/server.rs

Line 1578 in 7253be3

let cache = std::env::var("HUGGINGFACE_HUB_CACHE")

If HUGGINGFACE_HUB_CACHE is not defined, it will default to Cache::default(), which is equal to ~/.cache/huggingface/.

The issue here is that it can happen that HUGGINGFACE_HUB_CACHE is not set while HF_HOME is. In that case, the default above completely ignores HF_HOME. An easy fix is to change the default from Cache::default() to Cache::from_env() (see the definition here), which will take HF_HOME into account and will fall back to Cache::default() anyway if HF_HOME is not defined.