The map canvas was being passed to html2canvas as the target canvas. This is incorrect, as we expect to draw the annotation to its own canvas then draw that to the overall screenshot canvas using the computed offset. This was breaking the map canvas in 2D (the library was resizing it and drawing to it), and making capture fail in 3D (html2canvas expects a 2D canvas).
In 3D, the tail for annotations currently outside the map view were being drawn in the top/left corner of the screenshot. This was due to not testing if the SVG element was visible prior to adding it to the screenshot.
html2canvasas the target canvas. This is incorrect, as we expect to draw the annotation to its own canvas then draw that to the overall screenshot canvas using the computed offset. This was breaking the map canvas in 2D (the library was resizing it and drawing to it), and making capture fail in 3D (html2canvasexpects a 2D canvas).Fixes #738.