P14: Add self-serve user API key onboarding for distributed proxy

[wauputr4]

Goal

Make the distributed proxy usable by a normal Mizan user: register/login, create an API key, discover available distributed models, and call the OpenAI-compatible gateway.

User Story

As a user, I can self-serve an API key from a self-hosted Mizan instance and use it with an OpenAI-compatible client without knowing which daemon host serves my request.

Scope

• Review existing auth/API-key endpoints and close gaps for a self-serve distributed workflow.
• Add/adjust docs and scripts for user onboarding.
• Ensure /v1/models can include daemon-backed model routes/capabilities.
• Ensure model route visibility does not leak host secrets or private daemon metadata.
• Provide a smoke script for user registration/login/API key creation/model call.

Acceptance Criteria

• A fresh user can register/login and create a virtual API key through documented API calls.
• User can list daemon-backed models via /v1/models.
• User can call daemon-backed model using only Mizan base URL and virtual API key.
• Docs include copy-paste curl examples.
• Smoke script validates user API key onboarding end-to-end.

[wauputr4]

Completed via merged PR #78 (f165175): self-serve register/login/API-key onboarding docs and smoke script are present, and /v1/models includes daemon-backed model capabilities without exposing daemon host/private metadata. PR CI passed, and the remaining daemon-backed chat call dependency was handled by the follow-up dispatch lifecycle PR #80.