[Fix] dots1 expert bias routing

[fjosw]

What does this PR do?

This is a follow up to #41647. I went over the model code again and noticed that the dots1 routing apparently also broke in #40132. dots1 seems to be very close in architecture to Deepseek V3 but not explcitly dependent. Not sure if there is a more elegant way to fix this?

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?

@ArthurZucker

[fjosw]

Reminder: I think the dots1 model is still broken on main @ArthurZucker, not sure if this PR is the best way to address this, though.

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: dots1

[vasqu]

LGTM, I double checked and this looks correct to me. The selection is affected by the e_score_bias but not the final logits that are then selected.

Imo, we could refactor this later on to be within the router instead! (including deepseek). cc @3outeille since you worked a lot on MoEs recently it might be interesting

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.