feat(model-settings): add image generation model category and configuration

[skyfireitdiy]

Summary

Adds image_generation as a supported model category and capability across the model settings UI, default model configuration, and category inference. This lets users assign a default image generation model and filter models by this capability.

Changes

• modelCategory.ts: infer image_generation from model names (dall-e, flux, sdxl, stable-diffusion, midjourney, imagen, etc.); map the category to the image_generation capability; resolve it in resolveModelCategory.
• types/index.ts: add image_generation to ModelCategory union and DefaultModelsConfig.image_generation field.
• DefaultModelConfig.tsx: add an "Image Generation" default model slot (mirrors the existing speech recognition slot).
• ModelSettingsPage.tsx: add image_generation to the model category picker, default-model summary, and capability filter.
• modelCategory.test.ts: add coverage for image_generation category → capability mapping and selection.

Verification

• modelCategory.ts logic verified via node (18 assertions pass, including image_generation mapping).
• Syntax-checked all 5 changed files with tsc.

Closes #3279