feat(google): add thought_signature support for Gemini 2.0+

[Gabbosaur]

Summary

Add support for thought_signature in Google/Gemini client to enable function calling with Gemini 2.0+ models.

Problem

Gemini 2.0 and newer models (including gemini-2.0-flash) require thought_signature to be preserved and returned during multi-turn function calling conversations. Without this, the API returns:

400 INVALID_ARGUMENT: Function call is missing a thought_signature in functionCall parts.

Solution

1. FunctionCallBlock - Added optional thought_signature field
2. GoogleClient._response_to_client_response - Extract thought_signature from response parts
3. GoogleMemoryAdapter._turn_to_message - Include thought_signature when converting back to API format

Changes

• datapizza-ai-core/datapizza/type/type.py
• datapizza-ai-clients-google/datapizza/clients/google/google_client.py
• datapizza-ai-clients-google/datapizza/clients/google/memory_adapter.py

References

• Google AI - Thought Signatures Documentation

Testing

Tested with gemini-2.0-flash model using multi-turn function calling.

[Federaffo]

Thanks for the contribution. The PR currently fails CI due to formatting only: (blank line contains whitespace) in

Can you run the repo pre-commit pipeline locally and push the formatting fix?

Suggested commands:

• uvx ruff check . --fix
• pre-commit run --all-files

[Federaffo]

Thanks for the contribution. The PR currently fails CI due to formatting only:\n\n- W293 (blank line contains whitespace) in datapizza-ai-clients/datapizza-ai-clients-google/datapizza/clients/google/google_client.py:620\n\nCan you run the repo pre-commit pipeline locally and push the formatting fix?\nSuggested commands:\n- uvx ruff check . --fix\n- pre-commit run --all-files

Dismissing accidental duplicate review caused by command formatting.