kpoint density as input option for inpgen
It might be nice to have a 'kden' attribute ([1/Angstrom], or [1/abohr] or tuple of 3 densities) for the kpt namelist of the inpgen input (and maybe also to the fleur inp.xml as well)
In this mode inpgen looks at the reciprocal lattice vectors and estimates how many kpoints it needs in that dimension to be smaller that the given k-point density. And then just executes as usual for the estimated kpoint mesh (div1, div2, div3).
Since one is usually interested to converge with respect to the kpoint density, this is a helpful feature, also because this input value would be independent of the given crystal structure.