Mask out TetraSpeck beads with nuclear regions

[vreuter]

@TLSteinacker

We will lose some extra beads by extending a 2D nuclear mask throughout the full depth / along the full z-axis, but no big deal since usually we have plenty of legitimate bead spots. Furthermore, extending the $xy$ mask to 3D by simply translating it along $z$ prevents us from biasing our pool of beads, from which we select for drift correction and accuracy assessment, towards regions in which $z$ slices are disproportionately not occupied by a nucleus.

Needed to prevent mistaking a spot from FISH or IF for a fiducial bead