Improve discoverability of available algorithms for `reduce_dimensions`

Currently, users must leave the notebook or console and consult the documentation to discover which dimensionality reduction algorithms are supported (e.g., "umap", "pca", "tsne"). We should make the available algorithms discoverable directly from the notebook/code. Hyrax has functions for discovering available verbs, datasets, and models. Dimensionality reduction algorithms should provide a similar experience. Some ideas may be:

- Add CLI help. In `setup_parser` for `reduce_dimensions`, we can read the registered algorithms from the registry and include them in the `help` text.
- Add a function like `all_algorithm()` inside the `algorithm_registry.py`, similar to what verbs have.

Start in the `setup_parser` implementation for `reduce_dimensions` and inspect `algorithm_registry.py`, including how Hyrax discovers available verbs, datasets, and models. Determine whether the registered algorithms should be exposed through CLI help, an `all_algorithm()`-style function, or both. Done means users can discover supported algorithms such as `umap`, `pca`, and `tsne` directly from the notebook or code without consulting external documentation.

Written by the indexing model from the issue text.