Clean interface improvements

[krober10nd]

• this is a breaking change that will need some time and extensive testing to land
• addresses #180
• msh.clean()'s passive mode does not apply nodal valency reduction now.
• TODO: better handling of nodal attribute migration when applying moderate and aggressive cleaning.

[krober10nd]

If new nodes are added (such as is the case when running the nodal valency reducer) then this section is problematic for mapping nodal attributes (the intersection) as this will only map nodes that existed in the old mesh onto the new mesh. Thus, new nodes added in the cleaning operations will get default values.

            % f13
            if ~isempty(obj.f13)
                obj.f13.NumOfNodes = length(ind);
                for att = 1:obj.f13.nAttr
                    % Get the old index for this attribute
                    idx_old = m_old.f13.userval.Atr(att).Val(1,:);
                    val_old = m_old.f13.userval.Atr(att).Val(2:end,:);
                    % Only keep idx and val that is common to ind and map to ind
                    [~,ind_new,idx_new] = intersect(idx_old,ind);
                    val_new = val_old(:,ind_new);
                    % Put the uservalues back into f13 struct
                    obj.f13.userval.Atr(att).Val = [idx_new'; val_new];
                    obj.f13.userval.Atr(att).usernumnodes = length(idx_new);
                end
            end

[krober10nd]

Okay, this is now working on my end with running bound_con_init. In general, map_mesh_properties can now support the case of deletion and addition of new nodes. For the latter case, it uses nearest neighbor interpolation to put values in for those nodal attributes.

In summary, all levels of mesh cleaning can be used with the migration of the nodal attributes.

[krober10nd]

I think we can also use the new mapper in the BoundCr routine.

I'm on the fence about that because some nodal attributes depend on the bathymetry (e.g., primitive weights) and the bathy. slope should be recalculated if the nodes change. Also, the migration of boundaries is non-trivial and often has to be re-calculated anyway.

I'd rather the more advanced user do it afterwards from the old mesh to the new updated mesh via

m_old = map_mesh_properties(m_new,'msh_old',m_new); 

I think the application of nearest node mapping is suitable for the clean method however because this operation is often performed to clean up mesh quality or make boundaries traversable after the model is incrementally modified for instance. In general, for small changes.

[WPringle]

Fair enough. Yeah I think an obvious problem with using the map_mesh_properties as defined now after you did the BoundCr is that the bathymetry is already updated and there is no option to disable that mapping over at present. Let's just merge this in as you have it now.