LocalZProjector for large images opened as virtual stacks in Fiji.

We now are able to address large images that do not fit in RAM.

Fiji offers a simple solution for such a large images: open them as virtual stacks. In this case, each slice of the data is read from disk, on request. This PR ships a new version of the LZP plugin that exploit this, and generate and projection in two pass, exploiting the fact that we can process slice-by-slice.

It works but we needed to make sacrifices:

• you cannot project with median anymore.

Otherwise it works great.

src/main/java/fr/pasteur/iah/localzprojector/binning/BinningOp.java

@@ -72,7 +72,7 @@ public class BinningOp< T extends RealType< T > > extends AbstractUnaryFunctionO
 		ops().run( ChunkerOp.class, new CursorBasedChunk()
 		{
 			@Override
-			public void execute( final int startIndex, final int stepSize, final int numSteps )
+			public void execute( final long startIndex, final long stepSize, final long numSteps )
 			{
 				final RandomAccess< Neighborhood< T > > ra = ran.randomAccess( input );
 				final long[] currPos = new long[ input.numDimensions() ];