Make interp_points robust to non-equal grids and impractical default SciPy arguments

[rhugonnet]

This PR:

• Makes the behaviour of interp_points consistent between equal and regular grid by using both map_coordinates and interpn from SciPy depending on the grid,
• Removes the default prefilter argument of map_coordinates that created large data gaps,
• Corrects the method argument for map_coordinates by mapping it into an order argument,
• Adds detailed tests with synthetic data to ensure all results are consistent, in addition to the ones existing in test_xy2ij.

Resolves #483

TO-DO:

• Test the consistency of the AREA_OR_POINT argument between the methods.

[rhugonnet]

Merging this bug fix/improvement as it has pretty serious impact on xDEM, can be reviewed a posteriori.