[cairo] surface information saving with cairo_save

[Carl Worth]

On Mar 1, Thomas Hunger wrote:
 > I wonder if it would make sense to save the current surface matrix of the 
 > gstate->source if source is a pattern whenever one does a
 > cairo_save.

Yes, I think the attributes of the current pattern should be
considered a part of the current state and should be saved/restored.

So, we should fix this.

-Carl