Refactoring of DiscreteGridField and add ScalarFieldMapping

[damienmarchal]

DiscreteGridField is a component made a long time ago to cut dependencies between SofaImplicitField & SofaDistanceGrid.

The current implementation has defects, one of them is that the component can only load the grid content from a MHD file. So i refactor it to become a component with the sole purpose of storing a grid and doing field/grad/hessian evaluation and delegating to other component the content generation (from a file, from a field: see example-grid-generation-from-field.py). I also added experiment on using the grid that results in a mapping (see example scalar-field-mapping.py)

The PR:

• add const qualifier on the position parameter for every getValue/getGradient/getHessian
• move the MHD loader from the DiscreteGridField in a dedicated component
• add a GridSampler that can initialize the content of the DiscreteGridField from and ScalarField.
• add a ScalarFieldMapping the map from a positional 3D mstate to a 1D mstate (to be used in constraints)