Add type annotation and kwargs to run.io_bound and run.cpu_bound

[falkoschindler]

As noticed by @denniswittich, IO- and CPU-bound functions in RoSys are missing type annotations for parameters and return values. Apart from that run.cpu_bound is missing kwargs.

While this PR adds both, it unveils several type errors in connection with run.cpu_bound and run.io_bound:

rosys/run.py:26: error: Missing return statement  [return]
rosys/run.py:28: error: Return value expected  [return-value]
rosys/run.py:46: error: Missing return statement  [return]
rosys/run.py:48: error: Return value expected  [return-value]
rosys/vision/usb_camera/usb_device.py:85: error: Incompatible types in assignment (expression has type "Awaitable[Any] | None", variable has type "tuple[bool, Mat | ndarray[Any, dtype[integer[Any] | floating[Any]]]]")  [assignment]
rosys/vision/usb_camera/usb_camera.py:92: error: Argument 2 to "io_bound" has incompatible type "ndarray[Any, Any]"; expected "list[ndarray[Any, Any]]"  [arg-type]
rosys/vision/usb_camera/usb_camera.py:94: error: Argument 1 to "cpu_bound" has incompatible type "Callable[[bytes, ImageRotation, Rectangle | None], bytes | None]"; expected "Callable[[bytes, ImageRotation, Rectangle | None], bytes]"  [arg-type]
rosys/vision/mjpeg_camera/mjpeg_camera.py:94: error: Argument 1 to "cpu_bound" has incompatible type "Callable[[bytes, ImageRotation, Rectangle | None], bytes | None]"; expected "Callable[[bytes, ImageRotation, Rectangle | None], bytes]"  [arg-type]

We should fix them before merging into main.

[falkoschindler]

Ok, now there's only one mypy issue left related to an ndarray in usb_camera.py.
@NiklasNeugebauer Maybe you can have a look. 🙏

[NiklasNeugebauer]

@falkoschindler I was unable to figure out which type the result of capture.read might have in this case.
I doubt it is a list, but it might be an array with an extra dimension that is removed here by calling [0].
The related code section for handling raw MJPEG images is not used right now since the initial parameter checks are never executed. I was looking a little into the UsbDevice yesterday. It needs work at a lot of places, but nothing really critical, so maybe we can just talk about it in the next weekly.

For now, I decided to keep the current behavior, but execute it in the device and then deliver the raw bytes from there.