fix(tools): raise clear error for duplicate tool names in tool_runner

[Zawwarsami16]

I was poking around tool_runner and noticed that the runner accepts duplicate tool names without complaining. If you pass two tools that happen to share a name (say, two different get_weather impls imported from different modules), the second silently overwrites the first in _tools_by_name = {tool.name: tool for tool in tools}, and your first tool is gone before the request is even built. The API would reject the duplicate, but only with a generic 400 after the round trip, and by then you have no idea which side of your code was wrong.

This patch adds an upfront check in BaseToolRunner.__init__ that walks the tools, collects any names that show up more than once, and raises a ValueError listing them. It runs before any HTTP work, so you get the error at the call site of tool_runner(...) instead of inside a request handler. Behaviour for unique names is unchanged. I left the _tools_by_name dict comprehension in place and just added the validation pass above it.

Three new tests in tests/lib/tools/test_runners.py cover the sync runner, the async runner, and the case where multiple distinct names each have duplicates (the error should list all of them). uv run pytest tests/lib/ -n auto reports 496 passed, 1 skipped, 1 xfailed, and uv run ruff check . && uv run pyright && uv run mypy . are clean.