Add peak labels to free energy plots

[SCiarella]

This branch covers #203.
I have used a KD-tree to precompute a dictionary that maps each site to the closest peak.
Then after finding the path, I implemented a method to use the dictionary to identify which peaks get explored.
This information gets automatically plotted on the x-axis of the energy landscape:

[stefsmeets]

To my understanding the goal for the related issue is to link the coordinates back to the site labels from the cif file.

i.e. 48h instead of (28, 15, 10)

[SCiarella]

Now the path is compared to the real structure obtained from load_known_material(...) instead of the peaks of the density.
The figure have been updated in order to plot the site label on the top, and its Cartesian coordinates on the bottom:

[stefsmeets]

Hi @SCiarella , thanks for this PR. I just had a look and left some feedback. The plots look nice, though I'm wondering if it makes sense that the diffusing atom passes through so many cages so quickly.

My overall feeling is that this PR is not complete yet, and I'm not quite sure how this is supposed to be used.

• Can you add some tests for the plot (see tests/integration/plot_test.py)?
• Can you add some instructions for how this code is supposed to be used?
• The kdtree looks for the nearest atom for every voxel, shouldn't there be based some sort of distance threshold?

[SCiarella]

As discussed during the meeting on 09/01/24, I have implemented a color-coding to easily distinguish the closest site to the path:

In this last series of commits, I have also added a test for this plotting function.

[stefsmeets]

I'll review this next week when I'm back!

[stefsmeets]

I don't understand what is going on in this PR, it also includes the changes from #239, which makes this difficult to review. I tried to rebase to main to get rid of the already merged PR, but this did not work.

My main concern is that it is still not clear to me how this is supposed to be used. This api does not make sense to me. How would a user know they would need to run this?

structure = load_known_material('argyrodite')
vol.nearest_structure_reference(structure)
path.cartesian_path(vol)
path.fractional_path(vol)
path.path_over_structure(structure, vol)
plots.energy_along_path(path=path)

vs.

plots.energy_along_path(path=path)

What is the difference?

This is of course also a bit tied to #186. I'm also happy to have a go at updating the code. Otherwise let me know if you want to have a short chat about it!

[SCiarella]

Notice that I have not implemented the @cache decorator for fractional_path, cartesian_path and nearest_structure_reference.
The implementation is not trivial because our custom Volume and Pathway classes are not hashable.