fix(models): stream LiteLlm function-call args instead of buffering

[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: PROGRESSIVE_SSE_STREAMING is not honored by the LiteLlm adapter — function-call argument deltas are buffered until finish_reason #5342

Problem:
LiteLlm's FunctionChunk branch appends to args_parts and only yields on finish_reason. On b0180620, test_streaming_tool_call_args_assembled_from_many_fragments passed with len(responses) == 1, and grep PROGRESSIVE_SSE_STREAMING src/google/adk/models/lite_llm.py was empty. Long tool-call args do not stream.

Solution:
When PROGRESSIVE_SSE_STREAMING is on, yield a partial=True LlmResponse after each FunctionChunk. The FunctionCall uses will_continue=True and partial_args=[PartialArg(string_value=chunk.args)], where string_value is a delta (same as interactions_utils and the Gemini path). No json.loads on the partial path. _finalize_tool_call_response stays the last event. Flag off keeps today's single event.

Did not move LiteLlm onto StreamingResponseAggregator. It is Gemini-shaped (json_path on each PartialArg) and the change would be much larger.

Testing Plan

Unit Tests:

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

New tests in tests/unittests/models/test_litellm.py: test_streaming_function_chunks_yield_partials_when_progressive_sse_on and test_streaming_function_chunks_stay_buffered_when_progressive_sse_off. test_generate_content_async_stream_tool_call_includes_aggregated_text now pins the flag on and expects two function-call partials plus the aggregated event.

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

426 passed.

Reverted lite_llm.py to b0180620 and reran. The flag-on test failed (0 == 3 partials) and test_generate_content_async_stream_tool_call_includes_aggregated_text failed (4 == 6 events). The flag-off test still passed. pytest tests/unittests/{models,a2a,sessions,artifacts} on the fix: 2905 passed, 49 skipped, 3 xfailed.

Manual End-to-End (E2E) Tests:

Not run. No live LiteLLM provider key. The flag-on unit test covers a mock stream of three FunctionChunks plus finish_reason=tool_calls.

Additional context

Claim: #5342 (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.