Feature request: removal of invisible raster on export to reduce size

[Simon]

Example scenario: Import large images (png) into a svg file
and clip them to show only small interesting parts. The svg naturally
also includes true vector objects. When you export this into
pdf (or other formats that retain the vector nature of the figure),
the resulting file is gigantic as the entire images are saved. I have
seen overheads of up to 60MB (I work on scientific figures that should
be at least 600dpi). The entire raster appears to be saved, although the
inkscape knows about the bounding box of the clipping path. The images
can of course be cropped by hand with external tools, but it is
a tedious process, which must also be repeated if any images are replaced
with a new version.

Proposal: inkscape could use the bounding box of a clipping path
to determine the part of the image that is visible. It could add a few pixels
in each direction to avoid artifacts in antialiased render. This could
be an optional feature, selectable from the export dialog for pdf, eps
and other vector formats.

This transformation is completely lossless for png images, while for jpeg,
the recoding could introduce artifacts (can be avoided, if the clipping
corner is aligned to the jpeg block size).

The cropping can be done with calls to the imagemagick library.