Align client.images.generate() with OpenAI-compatible response schema

[flyworker]

Context

The current ImagesResource.generate() uses a custom request/response schema. mega-nova-api is standardizing on OpenAI-compatible format (MeganovaAI/mega-nova-api#434). The SDK should align.

Current State

• Request params: prompt, model, width, height, num_steps, guidance_scale, seed
• Response: ImageGenerationResponse(data: List[GeneratedImage]) where GeneratedImage(b64_json, revised_prompt)

Required Changes

• Add OpenAI-standard parameters: size (e.g. '1024x1024'), n, response_format ('url' or 'b64_json'), quality, style
• Keep backward-compat with custom params (width/height/num_steps) as aliases
• Support response_format='url' (return URL instead of base64)
• Align response model with OpenAI:

  class ImageData(BaseModel):
      url: Optional[str]
      b64_json: Optional[str]
      revised_prompt: Optional[str]

• Update example: examples/image_generation.py

Backward Compatibility

Keep existing params working — width=1024, height=1024 maps to size='1024x1024' internally.

[flyworker]

📋 Reference Report: AI API Compatibility Best Practices Research (29 items, 7 categories — protocols, gateways, SDKs, cloud platforms, local inference)

Tracking Epic: MeganovaAI/mega-nova-api#447

[flyworker]

Implemented in PR #15