Feature/hdbscan change

[Jacks0nJ]

Changed HDBSCAN to use scikit-learn's version.

Previously the HDBSCAN clustering algorithm was implemented using a mostly redundant package (simply called HDBSCAN), which has now been incorporated into scikit-learn. By making this change, less packages are needed, reducing the chance of dependency conflicts. E.g. the original HDBSCAN has issues with the version of numpy and compiling its own Cython code.

As the scikit-learn's HDBSCAN uses the same input format as the other algorithms in the clustering module, some of the unit tests have been changed. The most significant change is cluster _persistence is no longer available.

Both manual and unit tests have shown minimal changes to the clusters found. Only in one case did the unit test need to change, as the cluster labelling had changed for the same clusters found.

Finally, the condensed_tree_.plot is no longer available. Therefore the notebook How_Clustering_Works.ipynb was changed to use static .png images to explain the HDBSCAN algorithm. Some minor typos were also fixed.

[CKrawczyk]

All changes look good, feel free to merge when you are ready. A note about the removal of cluster persistence will need to be included in the release notes.

Further testing needs to be done to make sure the docker containers (now using a newer Python) are working as expected. These tests can be done on the update-dependencies branch. We should also check if Python versions 3.12 and 3.13 work yet. It might be worth bumping the docker containers to 3.13 if it does, that will give the longest time before it needs to be updated again.

The release notes will also need to address the new supported Python versions.