Improve Performance

[0lru]

Some operations are heavy-weight. As an example, "Canvas::drawPicture" may take some time.
Can we refactor this to release the GIL? That'd greatly improve the overall performance, I think.
Example of the pybind11-documentation:

m.def("call_go", &call_go, py::call_guard<py::gil_scoped_release>());