Model checks now report results in the model list

What changed

Provider health and connectivity checks are now a single Model Check action. Its dialog can check one model or all supported models; all-model results and errors appear beside each model instead of in a separate results drawer.

Why this matters to the user

Users can keep browsing, searching, and filtering models while a full check runs, and can inspect the result for each API key directly from a model row. Image, video, and audio generation models plus speech-to-text and text-to-speech models are skipped to avoid unnecessary generation charges.

What the user should do

Nothing — automatic. Use Model Check in Settings → Model Providers and select either a single model or all models.

Notes for release manager

Addresses issues #17935 and #18434. Failed API keys can be enabled or disabled from the result detail, and failed models can be removed after the check finishes.