feat: add corpus points to umap

[RogerHYang]

resolves #894
resolves #899
resolves #900
resolves #901

Screen.Recording.2023-07-12.at.8.49.12.AM.mov

[mikeldking]

I get your point that this is the shortest path to victory - but I do think it does muddy things, particularly from a domain modeling perspective. Can we file a ticket to properly delineate these types into the behavior we expect them to exhibit? I get that they roughly map to tables of sorts, but there's also bespoke behavior that is definitely a non-semantic ontology here.

[RogerHYang]

ok sounds good. to recap: Model is just some icing on a cake that's a table (in the sense of relational algebra), the trouble with corpus is that it has the cake but only half the icing, so it's much simpler to just keep all the icing on it, and maybe that's OK. Incidentally, what is a "Model" anyway if it mostly just walks and quacks like a table? Maybe what it needs is not surgery, but just a name change.

[mikeldking]

Approved - pending the outstanding comments

[RogerHYang]

had to restrict hdbscan version for it to compile on python 3.8

      Error compiling Cython file:
      ------------------------------------------------------------
      ...
      import numpy as np
      cimport numpy as np

      from libc.float cimport DBL_MAX

      from dist_metrics cimport DistanceMetric
      ^
      ------------------------------------------------------------

      hdbscan/_hdbscan_linkage.pyx:12:0: 'dist_metrics.pxd' not found

[mikeldking]

Failing build possibly due to scikit-learn-contrib/hdbscan#600

[mikeldking]

merging since scikitlearn hdbscan issues are un-related