replace TGVRegEnergyZV by h_zvl_energy
remove alpha, power_reg

change determinant regularizer to "beta * abs(1 / det + det)" (setting inf for det <= 0, causes problems in the optimization)

add determinant regularizer

BUGFIX: in LMSP-Method the effective penalty parameter is mu*template_tex_size/(rows*cols) instead of mu, this leads to a scaling of tau (the dual ascent step size)

take min_distance_quat_over_group as class member function to emphasize the asymmetry in quat1, quat2

fix sign in local rotation angle
add tau parameter (step size in ascent step) in multi level solver

fixed sampling and plot_transform_grid

bugfix: bool in cuda needs numpy.byte (see local_rotation flag)

add interpolation option for nearest neighbor interpolation in apply_transform
add several option in plot_transform_grid
bugfix: z-stride loop in apply_transform

add z-stride sampling loop in eval_f_sampling_cells, eval_f_u_sampling_cells, eval_f_z_sampling_cells

add automatic computation of reasonable block and grid dimensions

add grid_stride_kernel keyword in class KernelLaunch