Allow choice of color reduction algorithm for png8 images

Simone, you shared some palette generation code in google code right? What's the link?
Does that contain any usage of the neuquant algorithm shipped along with jai?

http://code.google.com/p/color-reduction-experiments/

it should test most of them. I am of course open to create new algorithms or to use third party APIs.

Made the issue title a bit less verbose

It seems this time I tried to bite more than I could chew.

Ok, so a little more research shows that:

• the original java neuquant algorithm does not deal with alpha channel in the generated palette
• the palette inverter code does not deal with alpha either

The code in the pngnq project deals with alpha, but has been severely reworked and thus requires some understanding of how the neural network works.
Finally both old and new Neuquant code assumes the image pixels are available in a flat array, and the code extracting it is definitely not memory efficient.

So I believe I'm going to put this on the back burner and try to solve one problem at a time, reading and understanding the neuquant article, adding proper unit testing around the code back ported from C (or I can take the Java code and pimp it up to 4 dims, but nevertheless, that requires time).

For the sake of keeping the history here is a output format I did some time ago, that dumps a png to file, runs pngnq on it, and returns the results. It's damn slow, result look very nice, and the palette does not really seem to work ok with alpha channel in.

Superceded by GEOS-4919

Switching all issues that have been in "resolved" state for more than one month without further comments to "closed" status