Various bugs with the merge tree states

[julien-tierny]

Describe the bug
Most of the merge tree states are affected by several bugs.

To Reproduce
In the ttk-data repository:
paraview states/mergeTreeClustering.pvsm
paraview states/mergeTreeFeatureTracking.pvsm
paraview states/mergeTreePGA.pvsm
paraview states/mergeTreeTemporalReduction.pvsm

Expected behavior
The planar layout of the trees includes spheres and tubes. The dimension reductions display properly, see the reference screenshots:
https://topology-tool-kit.github.io/img/gallery/mergeTreeClustering.jpg
https://topology-tool-kit.github.io/img/gallery/mergeTreeFeatureTracking.jpg
https://topology-tool-kit.github.io/img/gallery/mergeTreePGA.jpg
https://topology-tool-kit.github.io/img/gallery/mergeTreeTemporalReduction.jpg

Screenshots
Please find below screenshots which exemplify the issues on my system

Merge tree clustering
In the planar view of the trees:

• Spheres are missing for the nodes
• The tubes are not colored properly
  

Merge tree feature tracking
In the planar view of the trees:

• Spheres are missing for the nodes
• The tubes are not colored properly
  

Merge tree PGA
The 2D views are empty :(
(TTKMergeBlockTables2 is empty but ExtractSurface1 does contain the surface???!!!)

Merge tree temporal reduction
In the planar view of the trees:

• Spheres are missing for the nodes
• The tubes are not colored properly
  

System (please complete the following information):

• TTK dev
• ParaView 5.11
• numpy 1.25.2
• scikit-learn 1.1.2

Additional context
This bug is for @MatPont (after he's done with his manuscript write-up).

The color bug seems to be a consequence of one of Mathieu's latest PRs regarding the insertion of additional information along the arcs of the merge trees. Maybe it can be fixed in the state itself, but more likely with code adjustments.

The sphere, tubes and surface bugs are a little more surprising (these states are not tested by the CI because of their memory footprint). These should be fixable by state adjustments. Otherwise, one would need to roll back the history of commit to the last version of TTK for which these states were OK and search for the fault commits.

[julien-tierny]

Interestingly, these bugs do not seem to affect the PD-PGA state (topology-tool-kit/ttk-data#148).
Maybe the latest re-factoring could be the issue? (#944)

[MatPont]

Also interestingly, the states work fine on my computer with the current branches of ttk and ttk-data (except for the color bug in branches). However the libraries version differ.
Perhaps we would need to test these states on other computers (or different libraries version).

System
TTK dev
ParaView 5.10
Numpy 1.21.5
scikit-learn 0.23.2 (I'm not sure if we have checked the same thing here since it seems to be a big difference, I printed sklearn.__version__ after importing sklearn in python).

[MatPont]

Same screenshots after updating numpy to 1.25.2.
(It was just a quick attempt to reproduce the bug but I will investigate that, like you said, after the manuscript)

[julien-tierny]

I confirm that the issue with the missing geometry can no longer be reproduced (for mergeTreeClustering, mergeTreeTemporalReduction, mergeTreeFeatureTracking, mergeTreePGA).
It looks like this was a numpy related-issue.

[julien-tierny]

I confirm the issue is completely solved with PR #974 (tested on my home workstation).