Support `gpt-5.5` in built-in models, and preferably auto-discover upstream models instead of relying on a static local list.

[adndyxu]

Problem

Current client-visible models are generated from a hard-coded baseModels list, so newly available upstream models such as gpt-5.5 do not appear until Copool is updated.

Request

1. Add gpt-5.5 to the built-in models
2. Prefer fetching models from the configured upstream /v1/models
3. Fall back to the current static list if discovery fails
4. Keep Copool-specific reasoning aliases on top of discovered compatible models

Benefit

This would make Copool track upstream model availability automatically and reduce maintenance for every new gpt-5.x revision.

[adndyxu]

@AlickH @OrdinarySF 请问还维护吗？

[AlickH]

Implemented the minimal fix for this in v1.0.25 and shipped it in the latest macOS release.

Included in this release:

• added gpt-5.5 to the built-in client-visible model list
• updated the bundled proxyd model list and rebuilt the prebuilt binaries
• published a new notarized macOS release asset on GitHub Releases

Release: https://github.com/AlickH/Copool/releases/tag/v1.0.25

I did not implement upstream /v1/models auto-discovery in this pass. That part is still a separate follow-up if we want to replace the current static list.

[AlickH]

Follow-up: I also published a clean versioned release as v1.0.26, so the fix is now available under a new tag instead of only updating the existing v1.0.25 release.

Release: https://github.com/AlickH/Copool/releases/tag/v1.0.26