Edge case in LN2_PATCH_FLATTEN fills in zeros in the input values

[ofgulban]

I need to have a detailed look into this issue. E.g. if someone has patches of clusters within the multilaterate disk, the zeros get voronoi filled in. Workaround for this is to e.g. add 1 to the input values file, so that e.g. 1s and 2s become 2s and 3s and the zeros become 1, which is then correctly patch flattened and voronoi filled.