Feature request: investigate and support Standard/Fast speed selection for ChatGPT Codex

**What problem would this solve?**

Goose Desktop's ChatGPT Codex provider (chatgpt_codex, using ChatGPT sign-in) exposes reasoning effort but no Standard / Fast processing-speed choice. Users cannot request faster interactive responses while retaining their chosen model and reasoning effort. Lowering reasoning effort is not equivalent to Fast processing.

OpenAI documents Codex Fast mode: https://developers.openai.com/codex/speed

**What would a good outcome look like?**

- First verify whether Fast mode is supported through the authentication and backend route used by Goose's ChatGPT Codex provider.
- Where supported, expose Standard / Fast independently of reasoning effort, with Standard as the default and Fast explicitly opt-in.
- Respect account/model eligibility and explain unavailable or rejected requests. Do not imply Fast is active if it was not accepted.
- Explain additional credit/usage implications before enabling Fast, and define whether the selection persists per session or globally.
- If the route cannot support Fast, document the limitation and any supported alternative instead of exposing a nonfunctional toggle.

**Possible approaches**

Validate the official Codex client's wire-level behavior and server requirements first. The documented Codex configuration value `service_tier = "fast"` should not be assumed to be the identical or sufficient parameter for Goose's HTTP requests.

The chatgpt_codex request builder inspected during this investigation does not currently forward a service-tier selection. A Desktop selector alone would not suffice: provider request handling, configuration plumbing, UI, and tests should be considered together.

Verify eligible/ineligible accounts and models, unsupported/rejected requests, independence from reasoning effort, and avoidance of silent fallback presented as Fast. Distinguish ChatGPT subscription Fast mode from API-key Priority processing and their different billing models.

No Fast-mode requests were tested through Goose for this report. Server acceptance and eligibility remain unverified. This is a request to investigate feasibility and implement support where available, not a claim that every account/model can enable it with a client-only change.

**Additional context**

- Observed environment: Goose Desktop 1.49.0 on macOS; ChatGPT Codex provider.
- Searched open and closed issues/PRs for fast mode, service_tier, codex speed, and priority processing; no direct request was found.
- Distinct from GPT-5.6 Standard / Pro reasoning mode (#10514), Anthropic response metadata (#11630), and Goose's auxiliary fast-model path (#11337).

- [x] I have verified this does not duplicate an existing feature request

Do not begin implementation until the issue reaches **Ready** on the [Goose Issues board](https://github.com/orgs/aaif-goose/projects/1).

Commencez par examiner le générateur de requêtes du fournisseur chatgpt_codex dans la base de code Goose afin de comprendre la structure actuelle des requêtes HTTP. Consultez la documentation de l’API OpenAI Codex concernant le paramètre `service_tier` et testez l’éligibilité avec différents comptes et modèles. Déterminez comment intégrer une UI de sélection de la vitesse, mettre à jour la gestion de la configuration et gérer correctement les rejets ou les limitations du serveur.

Rédigé par le modèle d'indexation à partir du texte de l'issue.