Add line definitions to marching_triangles

[pjaap]

marching_triangles does now something similar to marching_tetrahedra:

It accepts lines and levels and returns now the points, adjecencies and values of a given function.

Changes in detail:

• inconsistencies with Tc, Tp and Ti parameters fixed
• a new method taking a lines vector added
• old methods are deprecated (TODO: what is the best way to do that?) and return only the point list, as before
• function values are interpolated linearly, as in marching_tetrahedra
• small performance caveat: the loop over levels is now one level higher and recomputes the point values every time; this makes implementation of lines easier. I think to slow down is reasonable

TODOs:
[] deprecate old methods correctly
[] deal with the new return values in calling functions (e.g., makie.jl)
[] if both lines and levels are given, the result is just a mess. But this happens too in marching_tetrahedra? Maybe throw a warning if both are non-empty.

[j-fu]

I think we don't need a very formal deprecation process here, just a breaking version bump
and a notice how to convert in the changelog. Two packages depend on it - GridVisualize and PlutoVista,
and I created GridVisualizeTools mainly in order to avoid a PlutoVista dependency on GridVisualize.

Before merging here, it would be good if you could make a PR to GridVisualize which shows that it works.
I would care about PlutoVista then. If this works as well, we could merge here.

[pjaap]

Ok, I pushed an update: the deprecated methods are now simply gone. In the changelog, there is a hint how to restore the old behavior, if desired.

The downstream PR is WIAS-PDELib/GridVisualize.jl#65

I'll fix the usage of marching_triangles there once this PR is merged.

[j-fu]

I think we should avoid O(N^2) here...

[j-fu]

Also, before merging, we should ensure that j-fu/PlutoVista.jl#16 works

[pjaap]

There is some nasty type conversion problem deep in j-fu/PlutoVista.jl#16 caused by this change. I am still figuring out how to get around this one.

[j-fu]

j-fu/PlutoVista.jl#16 works...

[j-fu]

Sorry, didn't read your previous post... I made a cheap workaround...
I have a general idea on this we could talk about, I guess this would be another PR...

[pjaap]

Oh, I now made a more sophisticated update here: Give the user control over the types of coordinates, values and points.
Then, in PlutoVista you can call

iso_pts = first(marching_triangles(pts, tris, f, [], collect(levels); Tc = Float32, Tv = Float32))

and

iso_pts0=first(marching_triangles(pts,tris,f,[],levels, Tv = Float32, Tc = Float32))

[pjaap]

With the two lines changed in PlutoVista test suites are fine in now in PlutoVista, GridVisualize and GridVisualizeTools

[pjaap]

@j-fu merge and register?

[j-fu]

In GridVisualize+PlutoVista, a minor version bump would be sufficient.