common: set Muse Glimmer thinking tags in common_chat_params

[paralin]

Overview

Set thinking_start_tag and thinking_end_tags for muse glimmer.

Fixes thinking / reasoning budget not being applied properly.

Additional information

   data.thinking_start_tag = " to=self<|message|>";
   data.thinking_end_tags  = { "<|eom|>" };

After the change, /v1/chat/completions with reasoning_budget_tokens: 32 and max_tokens: 384:

• finish_reason=stop (not length)
• reasoning contains the continuation message set in reasoning-budget-message

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: NO

[ggml-gh-bot]

Hi @paralin, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• PR Template not respected: Please respect the template when creating a new pull request. Make sure to fill out all required sections.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.

[pwilkin]

LGTM, @aldehir wanna take a peek?

[paralin]

@pwilkin My only question is if I should have defined it as a const string THINK_START? like in common_chat_params_init_cohere2moe - but this seems more straightforward

[paralin]

Double-checked and it works, before the commit, it would not respect reasoning budget, now:

• finish_reason=stop
• reasoning contains the budget continuation
• final content is the inspection plan
• 356 completion tokens, not a 384-token reasoning stall