Tool.from_tool() previously required a Tool object as its first argument, but the @tool decorator (since v3.0) returns the original function with metadata attached rather than a Tool object. This meant the documented workflow of decorating a function and then transforming it before registration would crash with AttributeError: 'function' object has no attribute 'parameters'.

Now from_tool accepts either a Tool or a callable, auto-coercing the latter via Tool._ensure_tool(). This respects any @tool decorator metadata (name, description, etc.) on the function, so the documented pattern works naturally:

@tool
def search(q: str, limit: int = 10) -> list[str]:
    """Search for items."""
    return [f"Result {i} for {q}" for i in range(limit)]

better_search = Tool.from_tool(
    search,
    name="find_items",
    transform_args={"q": ArgTransform(name="query")},
)

mcp = FastMCP("Server")
mcp.add_tool(better_search)

The coercion reuses FunctionTool.from_function(fn, metadata=meta) rather than expanding kwargs, keeping it in sync with the existing metadata path.

Closes #3232