feat: add GitGot as a model provider

[respectus]

What

Adds GitGot as a model provider. It is an OpenAI-compatible endpoint, so core/llm/llms/GitGot.ts is a thin OpenAI subclass with a base URL — the same shape as DeepInfra.ts and the other hosted OpenAI-compatible providers — plus the usual registrations in autodetect.ts, openai-adapters, config-types, the VS Code config schema, and the Add-New-Model UI.

Why

It serves open-weight models that have few other providers:

model	context
moonshotai/Kimi-K2.7-Code	262,144
moonshotai/Kimi-K2.6	262,144
deepseek-ai/DeepSeek-V4-Pro	163,840
deepseek-ai/DeepSeek-V4-Flash	163,840
openai/gpt-oss-120b	131,072
meta-llama/Llama-3.3-70B-Instruct	131,072

Kimi K2.7 Code in particular is a code-tuned model with a 262k context, which is the case for it in this extension specifically.

Model packages

gptOss120B and llama3370bChat already exist, so those gain a GitGot dimension rather than a duplicate entry. kimiK27Code and deepseekV4Flash are new.

Reasoning controls

The docs page is explicit that reasoning_effort behaves differently per model on this host, because the models do:

• openai/gpt-oss-120b — graded. Mean reasoning_content over three runs: low 283 chars, medium 469, high 1289.
• DeepSeek V4 — an on/off switch. With the field absent, reasoning_content was empty on every run; with it present at any value, reasoning ran, and the value made no monotonic difference.
• Kimi — reasoning runs regardless and nothing the caller sends changes it.

All of them return reasoning in reasoning_content. I mention it because a caller who assumes one shape across all six gets a control that silently does nothing.

Tests

Added to the existing createOpenAISubclassTests suite rather than a new file:

npx vitest run llm/llms/OpenAI-compatible.vitest.ts
✓ 218 tests passed, including 7 for gitgot

tsc --noEmit: core, packages/openai-adapters, packages/config-types — all clean
prettier --check on every touched file — clean

The models list with live prices and context limits is at https://inference.gitgot.ai/v1/models, no key needed, if it is useful for verifying any of the above.

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}