Re-add MD-TASK to project, apply clustering on the correlation matrix

[AndrewRadev]

The atomic movement similarity matrix from GeoStaS might produce some weird results due to its attempts to detect rotationally-rigid domains. It might be useful to explore movement correlation via MD-TASK (https://md-task.readthedocs.io/en/latest), maybe clustered via agglomerative clustering with a constraint for physically-close residues.