feat(mcp): propagate grounding metadata from MCP _meta

[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: Support Grounding Metadata / Citation Propagation in in mcp_tool class #6081

Problem:
_maybe_add_grounding_metadata returns early unless a tool is named google_search_agent. I ran _handle_after_model_callback with a dummy tool and temp:_adk_grounding_metadata={'foo': 'bar'} set: the callback returned None and LlmResponse.grounding_metadata stayed unset. git log -S propagate_grounding_metadata -- src/google/adk/flows/llm_flows/base_llm_flow.py shows d689a04f replaced the name check with getattr(tool, 'propagate_grounding_metadata', False), and b2daf83d put the name check back. McpTool._run_async_impl dumps CallToolResult and never writes the temp key.

Solution:
Restore the d689a04f gate: attach when some canonical tool has propagate_grounding_metadata and the temp key is set. Keep the canonical_tools_cache read. Add propagate_grounding_metadata: bool = False on McpTool and McpToolset. When True, validate adk_grounding_metadata from dumped meta / _meta as GroundingMetadata and write the temp key. The tool return dict is unchanged. A ValidationError is logged and skipped.

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/flows/llm_flows/test_base_llm_flow.py::test_handle_after_model_callback_grounding_with_no_callbacks tests/unittests/flows/llm_flows/test_base_llm_flow.py::test_handle_after_model_callback_grounding_with_callback_override tests/unittests/flows/llm_flows/test_base_llm_flow.py::test_handle_after_model_callback_grounding_with_plugin_override tests/unittests/flows/llm_flows/test_base_llm_flow.py::test_handle_after_model_callback_caches_canonical_tools tests/unittests/tools/mcp_tool/test_mcp_tool.py::TestMCPTool::test_run_async_impl_propagates_grounding_metadata_from_meta tests/unittests/tools/mcp_tool/test_mcp_tool.py::TestMCPTool::test_run_async_impl_skips_grounding_metadata_when_flag_off

15 passed.

The cache test uses a tool named research_agent with propagate_grounding_metadata=True. With the three src files checked out from upstream/main, that test fails with AttributeError: 'NoneType' object has no attribute 'grounding_metadata', and the MCP flag-on test raises TypeError: McpTool.__init__() got an unexpected keyword argument 'propagate_grounding_metadata'.

Manual End-to-End (E2E) Tests:

Not run. No live MCP server or Gemini Enterprise UI on this machine.

Additional context

Claim: #6081 (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.