test(google_genai): migrate golden coverage to vcr tests (#196)

Move the remaining Google GenAI scenarios that were only exercised by the manual golden suite into the integration test module with recorded VCR cassettes.

The new coverage focuses on image and document inputs, prefill, short max-token truncation, tool-result follow-ups, and reasoning metrics while avoiding duplicate tests for flows already covered in the integration suite.

Remove the now-redundant genai golden suite and trim the golden README so it only describes the remaining suites.

Author: AbhiPrasad

internal/golden/README.md

@@ -17,32 +17,32 @@ Each test suite validates:
 
 ## Running Tests
 
-### Python Tests
+Run a specific golden suite from its directory, for example:
 
 ```bash
-cd genai-py-v1
-python google_genai.py
+cd langchain-py-v1
+python langchain.py
 ```
 
-## Requirements
+```bash
+cd pydantic-ai-v1
+python pydantic_ai_test.py
+```
 
-Before running the tests, ensure you have the appropriate API keys set as environment variables:
 
-- `BRAINTRUST_API_KEY` to log to braintrust
-- `GOOGLE_API_KEY` for Google Generative AI tests
+## Requirements
 
-The tests will automatically log traces to Braintrust projects named `golden-python-genai`.
+Before running a suite, ensure you have the appropriate API keys set as environment variables for that provider, along with `BRAINTRUST_API_KEY` if you want to log traces to Braintrust.
 
 ## Contributing
 
 ### Adding a New Provider
 
 To add tests for a new AI provider:
 
-1. Use `google_genai.py` as a reference implementation
-2. Provide it as context to an LLM and ask it to create an equivalent file for the new provider
-3. Ensure all test cases are covered with provider-specific adaptations
-4. Follow the naming convention: `<provider>.py`
+1. Use an existing golden suite as a reference implementation
+2. Ensure all test cases are covered with provider-specific adaptations
+3. Follow the naming convention already used by the surrounding suites
 
 ### Adding New Feature Coverage
 
@@ -52,4 +52,4 @@ When adding a new feature (like reasoning, extended context, or new modalities):
 2. Ensure consistency in test structure and naming across providers
 3. Update this README to document the new feature coverage
 
-This ensures comprehensive validation across all supported providers and maintains test parity.
+This helps maintain broad feature coverage across the remaining golden suites.