fix(models): let LiteLlm pin the tool-result role

[claxman]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: [LiteLLM] role='tool_responses' for Gemma-4 models breaks tool calling on LM Studio (strict OpenAI role validation) #6482

Problem:
_content_to_message_param sets role tool_responses whenever the model name matches Gemma-4. On this tree, google/gemma-4-e4b yields tool_responses and openai/gpt-4o yields tool. grep tool_result_role src/google/adk/models/lite_llm.py was empty. #6482 reports that this breaks servers that validate OpenAI roles.

Solution:
LiteLlm(tool_result_role="tool") is a declared field that pins the role. Unset keeps Gemma-4 auto-detect. It is popped from _additional_args so LiteLLM does not see it. _resolve_tool_result_role is used by _content_to_message_param, _ensure_tool_results, and _get_completion_inputs.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

PYTHONPATH=src .venv/bin/python -m pytest -q tests/unittests/models/test_lite_llm_gemma_tool_role.py

14 passed.

Reverted lite_llm.py to b018062 and reran the file: 6 failed, 8 passed. Restored: 14 passed.

Manual End-to-End (E2E) Tests:

Not run here. No LM Studio on this machine. The LM Studio gemma-4-e4b repro from the issue with LiteLlm(model=..., tool_result_role="tool") would cover it if the reporter can run it.

Additional context

Claim: #6482 (comment)

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.