handle the caching case in setImagePath properly; this is a significant increase in complexity in this rather fundamental method.

in combination with the change in Svg::setImagePath to call FrameSvg::setImagePath directly (due to an API wart i only today noticed), this has the potential to cause new issues, though it is running quite nicely on my system.

needs extensive testing, but should also hopefully take care of the remaining crash being seen related to the FrameData cache