Add more options to spatialstats functions

[rhugonnet]

Some more flexibility for some spatialstats functions, without changing the default behaviour.

• fit_sum_model_variogram can now take an inital guess of variogram parameters that are later passed to scipy.curve_fit: p0 and bounds bounds
• plot_vgm can be passed xlabel, ylabel, xlim and ylim to more easily customize the plot (which can be hard with plt.gca()as the Figure is parted into subpanels within a grid)
• fixed a bug in patches_method to allow for a "only 100% valid data" when checking patches no-data with the input threshold

For plots: it would be good to have a consistent approach at plotting and passing axes/grid objects to allow easy user edits without duplicating input parameters. But, realistically, this would have to be done in a later PR focusing on a plotting library. (skgstat plotting functionalitiers allow to pass ax, grid, and fig object back and forth, maybe we can check that out? or maybe there's better solutions out there, it's quite a common functionality people want to have)

[rhugonnet]

@adehecq @erikmannerfelt
Ready to be merged, just a tiny bug fix and some more flexiblity in the fit/plot vgm functions.