Fix e2e tests by switching to native Gemini integration

[josemontespg]

Description

This PR fixes the failed e2e tests for restaurant_finder in CI.

Error

The tests failed with the following error from LiteLLM:

litellm.NotFoundError: Vertex_ai_betaException - b'{\\n  "error": {\\n    "code": 404,\\n    "message": "models/gemini-2.5-flash is not found for API version v1beta, or is not supported for generateContent. Call ListModels to see the list of available models and their supported methods.",\\n    "status": "NOT_FOUND"\\n  }\\n}\\n'

Fix

As suggested by the warning in the logs, I switched from using LiteLlm to native Gemini integration in RestaurantAgent (samples/agent/adk/restaurant_finder/agent.py).
This avoids the issue with LiteLLM trying to use an unavailable API version.

Verification

I verified that the tests pass locally after:

1. Switching Flutter to the beta channel (version 3.44.0-0.3.pre) to match CI.
2. Using gemini-3-flash-preview as the model (by setting MODEL_NAME environment variable), as gemini-2.5-flash continued to return 404 errors even with the native SDK.

All 4 tests passed locally.

[gemini-code-assist]

Code Review

This pull request migrates the restaurant finder agent from LiteLlm to the native Gemini model integration. The review feedback identifies a typo in the default model name (gemini-2.5-flash) and suggests a more robust approach for parsing model IDs from environment variables to accommodate various provider prefixes.