[Bot] Update Inference Providers documentation

[HuggingFaceInfra]

This PR automatically upgrades the @huggingface/tasks and @huggingface/inference packages and regenerates the Inference Providers documentation by running:

cd scripts/inference-providers
pnpm update @huggingface/tasks@latest @huggingface/inference@latest
pnpm run generate

This PR was automatically created (and merged) by the Update Inference Providers Documentation workflow.

---

Note

Low Risk
Documentation-only, generated markdown changes with no runtime or API behavior in this repo.

Overview
Regenerated Inference Providers docs after bumping @huggingface/tasks and @huggingface/inference, so example InferenceSnippet model IDs and recommended-model lists track current provider catalogs.

Cohere chat-completion examples now use CohereLabs/tiny-aya-global instead of tiny-aya-earth. HF Inference swaps doc examples to ibm-granite/granite-embedding-97m-multilingual-r2 (feature extraction) and deepset/minilm-uncased-squad2 (question answering). SambaNova no longer documents a Chat Completion (LLM) task block, and SambaNova is dropped from the multi-provider chat-completion LLM snippet. Chat completion and image-text-to-text pages add Qwen/Qwen2.5-VL-3B-Instruct as a recommended VLM.

^{Reviewed by Cursor Bugbot for commit 887affc. Bugbot is set up for automated code reviews on this repo. Configure here.}

[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.