fix: provide ToolContext for local tool hooks

[pragnyanramtha]

Summary

Provides ToolContext to lifecycle hooks for built-in local tool calls that have concrete tool call IDs.

Function/custom tools already expose context.tool_call_id through ToolContext, but built-in local tools such as LocalShellTool, ShellTool, ComputerTool, and ApplyPatchTool passed a plain RunContextWrapper to on_tool_start / on_tool_end. This made it difficult to correlate start/end hook calls for metrics when tool calls run concurrently.

This PR keeps public hook signatures unchanged and constructs a ToolContext only for hook callbacks. Executor request contexts remain the existing RunContextWrapper values.

Test plan

Passed:

• uv run pytest tests/test_computer_action.py tests/test_local_shell_tool.py tests/test_shell_tool.py tests/test_apply_patch_tool.py (67 passed)
• make format-check
• make lint
• uv run mypy src/agents/run_internal/tool_actions.py src/agents/lifecycle.py tests/test_computer_action.py tests/test_local_shell_tool.py tests/test_shell_tool.py tests/test_apply_patch_tool.py
• uv run pyright src/agents/run_internal/tool_actions.py src/agents/lifecycle.py tests/test_computer_action.py tests/test_local_shell_tool.py tests/test_shell_tool.py tests/test_apply_patch_tool.py
• git diff --check

Attempted repo-wide checks:

• make typecheck did not complete because the local environment is missing optional dependencies/stubs such as numpy, litellm, temporalio, sqlalchemy, and boto3, plus existing Python 3.11/runtime typing issues unrelated to this patch.
• make tests reached the broader suite but failed on missing optional extras/imports such as sqlalchemy, litellm, numpy, and runloop_api_client.

Issue number

Fixes #1849

Checks

• I've added new tests (if relevant)
• I've added/updated the relevant documentation
• I've run make lint and make format
• I've made sure tests pass

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.