feat: Add support for GenAI (Google) GoogleGenAIChatGenerator structured output

[aelhajj]

Is your feature request related to a problem? Please describe.

GoogleGenAIChatGenerator doesn't support structured outputs. Users coming from OpenAIChatGenerator expect generation_kwargs={"response_format": MyModel} to work, but it's silently ignored.

Describe the solution you'd like

Accept response_format in generation_kwargs (Pydantic model or JSON schema dict) and internally translate it to Google's native response_schema + response_mime_type: "application/json" in GenerateContentConfig. Should work for both run() and run_async().

Describe alternatives you've considered

• Passing response_schema + response_mime_type directly in generation_kwargs. Works but undocumented, requires Google SDK internals knowledge, and breaks cross-provider portability.
• Wrapping the component in a custom pre-processing layer. Adds boilerplate and puts the burden on users.

Additional context

• OpenAIChatGenerator, MistralChatGenerator, NvidiaChatGenerator and others already support response_format. GoogleGenAIChatGenerator is notably absent from [the structured output tutorial](https://haystack.deepset.ai/tutorials/28_structured_output_with_openai).
• Google's SDK docs on structured output: https://ai.google.dev/gemini-api/docs/structured-output
• The OpenAIChatGenerator structured output implementation (v2.18.0) can serve as reference.
• I'm willing to implement this, looking for maintainer alignment on the approach before opening a PR.