This improves the algorithm for suggesting and subsampling factors / chunk sizes:
We don't consider dimensions with size 1 for (further) downsampling
We use the dimensions of downsampled image in chunk size determination. For example we will never suggest chunk size [16,16,16] for 2D images. Knowing that dim[Z]==1, the cell size will be limited to 1 in Z.
This improves the algorithm for suggesting and subsampling factors / chunk sizes:
This should resolve https://github.com/bigdataviewer/bigdataviewer_fiji/issues/21