vout_subpictures: remove assert on text rendering

Text-to-speech renderers don't set a new chroma on the output since they don't create an actual picture.

The code before 7fef0ec0 was returning if the chroma wasn't changed, which is fine for Text-to-speech. There is no picture to place.

Also avoid returning errors when everything was fine, as in f0f1a500. But nsspeechsynthesizer doesn't return an error. It's would be more logical to check if there's any picture to display rather than pretending the rendering failed.