[FEA] Approximate_predict HDBSCAN Support

[whymauri]

Is your feature request related to a problem? Please describe.

• Support or HDBSCAN to store training datapoints for later approximate inference without re-fitting the underlying HDBSCAN model.
• For example, sklearn-contrib's HDBSCAN has the approximate_predict() method.
• References to previous requests: [FEA] Support prediction on new data with HDBSCAN.  #4472
• [FEA] Need a approximate_predict function for cuml HDBSCAN #4448

Describe the solution you'd like
Similar to scikit-learn HDBSCAN's approximate_predict
https://hdbscan.readthedocs.io/en/latest/api.html#hdbscan.prediction.approximate_predict

Predict the cluster label of new points. The returned labels will be those of the original clustering found by clusterer, and therefore are not (necessarily) the cluster labels that would be found by clustering the original data combined with points_to_predict, hence the ‘approximate’ label.

Describe alternatives you've considered
Forking CuML and building the functionality, but perhaps it's already on your roadmap?

I saw on one of the issues it was slated for 22.04, but the 22.04 issue/project was closed/resolved without comment. Is it planned for a future release? I could take a stab at implementation for a future release if the issue is lack of resources.

[divyegala]

@whymauri it was de-prioritized for 22.04. I will get back on this and put it on my priority plate.

[whymauri]

Thank you Divye, I appreciate the transparency :)

[divyegala]

@whymauri no worries! As always, we welcome all contributions since you mentioned in your alternatives that you could want to implement this feature. If you would like to do that, I can be a resource to you in any way that is needed to get the PR across the finish line.

[VibhuJawa]

Closing in favor of earlier issue: #4448