Align image pipes with latest provider features and OpenWebUI patterns

[yuzukumo]

Summary

This updates the image pipes to better match the latest official capabilities of the models we currently target in OpenWebUI.

The main focus of this pass was keeping the implementations clean and provider-aligned, while making sure they work properly with current OpenWebUI file handling.

Updated models

• Doubao: Seedream 5
• xAI: grok-imagine-image, grok-imagine-image-pro
• Gemini: gemini-3.1-flash-image-preview, gemini-3-pro-image-preview
• OpenAI: gpt-image-1.5

What changed

• cleaned up the provider-specific request and response handling for the current image models
• aligned image upload handling with current OpenWebUI helpers and reserved args
• tightened Gemini image support around the latest preview models
• kept OpenAI compatibility handling minimal and limited to the cases we actually needed in testing

Validation

• tested the related image pipes in real OpenWebUI usage
• ran make lint

[OrenZhang]

鉴于二开主仓库 OVINC-CN/OpenWebUI 短期内不会合并 OUI 的新版本更新内容，所以本次 PR 的变更也会暂时挂起