[2/2] feat: report whether the calling account can use each generation model

[dazzatronus]

Most models are now plan-gated, so a client that offers every model sends users to a 403. The models endpoints can report, per calling API key, which models the account's plan can use. This documents that: an optional available boolean and an unavailableReason enum on GenerationModel, plus a line on the list endpoint telling clients to offer only available models.

available is optional on purpose. When the account can't be checked the API omits it, and the description tells clients to treat a missing value as unknown, not as available. Generation requests are checked either way.

Verify: pnpm build and pnpm test pass on Node 22. Live /models, /models?expand=options and /models/{id} responses carrying available parse; so does an unavailable model with AiCapabilityNotIncluded and one with availability omitted. An unknown unavailableReason is rejected.

Stack

Order	PR	Adds	Why separate
1	#141	Declares name and description	Corrects the spec to the current response
2	#142 current	Adds available and unavailableReason	New field clients will rely on; revertible alone

Merge bottom-up: #141 into main, then retarget #142 to main.

flowchart LR
  A["Client lists generation models<br/>with its API key"] --> B{"Account's plan includes<br/>this kind of generation?"}
  B -- yes --> C["available: true"]
  B -- no --> D["available: false<br/>+ unavailableReason"]
  B -- "can't confirm" --> E["available omitted"]
  C --> F["Offer the model"]
  D --> G["Hide or disable the model"]
  E --> H["Treat as unknown;<br/>generation is still checked"]

Loading

Rules: available and unavailableReason, list endpoint guidance.

[github-actions]

🎉 This PR is included in version 1.19.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀