Add selectable OpenAI Responses API image generation mode

[tianheil3]

Summary

Adds a selectable OpenAI image generation API mode for direct API media generation:

• images: existing /v1/images/generations flow
• responses: new /v1/responses flow using the image_generation tool

This helps OpenAI-compatible gateways and newer image models that expose image generation through the Responses API rather than the legacy Images API.

Changes

• Adds api_providers.openai.image_api_mode configuration
• Adds an OpenAI image API selector in the Web UI settings
• Passes the selected mode through APIProviderMediaService and ImageClient
• Implements Responses API image generation in ImageGPT
• Supports extracting generated images from nested response payloads as URL or base64
• Keeps the existing Images API behavior as the default for backward compatibility

Verification

• Ran Python syntax checks on changed files:
  • pixelle_video/config/schema.py
  • pixelle_video/services/api_media.py
  • pixelle_video/services/api_services/config.py
  • pixelle_video/services/api_services/image_client.py
  • pixelle_video/services/api_services/image_gpt.py
  • web/components/settings.py
• Verified Responses API mode with an OpenAI-compatible gateway and gpt-image-2; generation returned a local PNG successfully.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}