updated docs of the new clustering models APIs

[MechaCritter]

Related Issues

Proposed Changes:

This brings the docs in line with the new encoder API. The encoders no longer take pre-fitted kmeans_model / gmm_model objects or lean on the pretrained KMeansWeights / GMMWeights enums as the primary path — they're now configured from parameters (n_clusters / n_components plus kmeans_params / gmm_params / pca_params), trained with learn(), and persisted with save_to_disk / load_from_disk. The old weight-enum path still works but is deprecated, and the docs needed to say so.

The main changes:

• New docs/clustering/README.md. Documents the pyvisim.clustering package (KMeans, GaussianMixtureModel, PCA) that the encoders now build their vocabulary from — what each model exposes, the diagonal-covariance constraint on the GMM, and the internal _from_sklearn adoption path the legacy weights use.
• Reworked encoder construction docs. base_encoder.md now describes the params-based constructors and the train/save/load lifecycle instead of the old two-ways-to-supply-a-model story. Added "Constructing one" sections with runnable snippets to vlad.md and fisher_vector.md.
• Marked the weights path deprecated. weights.md gets a deprecation banner up top and points readers at the learn() + save_to_disk() replacement. Also fixed a couple of now-stale internals (the PCA is loaded by _load_pretrained_weights, not ImageEncoderBase.__init__).
• Overview updates. overview.md lists the new clustering/ area and replaces the "pretrained weights are enums" bullet with the .encoder persistence story.

I kept the deprecated weights page around rather than deleting it, since the enum path still functions — it just warns now.

How did you test it?

Docs-only change, no code touched. Verified the relative links between the new and edited pages resolve, and that the referenced symbols (save_to_disk, load_from_disk, kmeans_params, gmm_params, pca_params, _from_sklearn, _load_pretrained_weights) match the current pyvisim source.

Notes for the reviewer

Worth a sanity check that the code snippets in vlad.md / fisher_vector.md match the actual constructor signatures, and that I've described the diagonal-covariance constraint and the "don't pass n_clusters / n_components in both places" ValueError behavior correctly.

Checklist

• I have read the contributors guidelines.
• I have updated the related issue with new insights and changes.
• I have added unit tests and updated the docstrings.
• I've used one of the conventional commit types for my PR title: docs:.
• I have documented my code.
• When applicable: I have reviewed the AI-generated code sections, according to contributors guidelines.
• I have run pre-commit hooks and fixed any issue.