[Inference Providers] drop the conversational tag requirement

[hanouticelina]

The Hub no longer requires the conversational tag to register a model as a chat-completion model for inference providers (internal PR: huggingface-internal/moon-landing#19696). New rule: any model with pipeline tag text-generation or image-text-to-text can be registered as conversational — the registered mapping decides, not the model's tags — and the legacy text-generation inference task is no longer accepted at registration.

This updates register-as-a-provider accordingly, and switches the "browse chat models" links from other=conversational to pipeline_tag=text-generation so models without the tag are included.

---

Note

Low Risk
Docs-only updates to links and provider registration guidance; no runtime or API code changes in this repo.

Overview
Documentation now matches Hub behavior where chat-completion registration no longer depends on the conversational model tag—providers register text-generation and image-text-to-text models with task conversational, and the mapping defines chat completion rather than requiring a conversational tag on the model.

register-as-a-provider.md rewrites the chatCompletion tip and Model Mapping API notes: validation is conversational for those pipeline tags, task equals pipeline_tag otherwise, and examples no longer imply tagging models as conversational.

Browse links in models-widgets.md, inference-providers/index.md, and integrations/codex.md switch from other=conversational to pipeline_tag=text-generation so model lists include chat-capable models without the legacy tag.

^{Reviewed by Cursor Bugbot for commit 8a42b91. 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.