readthedocs API changes

[jeremymanning]

I think a few minor changes would help make the API documentation a bit clearer on readthedocs:

• Make a note in hypertools.tools.describe_pca that the function is depreciated (replaced by hypertools.describe
• Change order of main functions to: plot, analyze, normalize, reduce, align, cluster, describe
• Make a note in hypertools.tools.procrustes that the function is depreciated (replaced by hypertools.align(..., align='procrustes')
• Does cluster support normalize/reduce/align keywords? I think it should... (if not, we should open another issue for it; otherwise we should correct the API)
• In plot, analyze, normalize, reduce, align, and cluster we should add more details on how to specify the models. Specifically, make clear that the user should pass in a string (list possible options) if they want the default parameters. For advanced users, they should pass a dictionary (and make the scikit-learn "supported models" a hyperlink, so advanced users can explore which parameters are supported)
• For DataGeometry in the high level description, instead of saying "Note: this class should not be called directly, ..." I think we should say "Note: new instances of this class should not be initialized directly, ..." (or something along those lines)

[jeremymanning]

Status in the 1.0 update (PR #272 #272): 4 of 6 asks here are already resolved, but not in the way originally requested — describe_pca and procrustes were removed outright (not deprecated-with-a-note), and DataGeometry's docstring was rewritten to mark it internal-only (stronger than the requested wording tweak). The model-spec string-vs-dict docs ask is genuinely done in reduce()/align()/cluster(). Two items are still open: function listing order, and whether cluster() should accept normalize/reduce/align keywords.

How this could be built on the 1.0 architecture:

• Reorder imports in hypertools/__init__.py (currently plot, load, analyze, reduce, align, normalize, describe, cluster) to the requested plot, analyze, normalize, reduce, align, cluster, describe, and mirror that order in docs/api.rst's autosummary blocks.
• For the cluster-composition ask: hypertools/tools/analyze.py's analyze() already composes normalize → reduce → align via aligner(reducer(normalizer(...))), but does not include cluster, and hypertools/cluster/cluster.py's cluster(x, cluster="KMeans", n_clusters=3, format_data=True) has no normalize/reduce/align kwargs. Two honest paths: (a) add those as pass-through kwargs to cluster() that call the same normalizer/reducer/aligner helpers analyze() uses before fitting (medium effort, real feature work), or (b) just document in cluster()'s docstring that this composition happens via hyp.plot()/hyp.analyze() today (docs-only, low effort).
• Swap the current prose ("See scikit-learn specific model docs...") for real sphinx cross-references/links to the relevant sklearn pages in reduce()/align()/cluster() docstrings.

Alternatives / prior art: n/a — this is a docs-organization ask, not a design question.

Note: parts of this ask are superseded — describe_pca/procrustes no longer exist to annotate (replaced by describe()/align(..., align='procrustes')), and DataGeometry is already documented as internal-only.

Tagging as low effort; leaving open for consideration in a future release.