docs: document spectral-line datacube modeling in the feature overview

Datacube modeling shipped in May 2026 (list of Interferometer datasets in an
af.FactorGraphModel, shared lens model, per-channel source, shared curvature
matrix across channels), but the docs never mentioned it — a user reading the
feature overview or the "What Dataset Type?" routing had no way to learn cubes
were supported.

- overview_3_features.md: add a "Data Cubes" one-line summary and a full
  section covering the list-of-datasets design, the shared-lens/per-channel
  -source split, and the channel-invariant curvature sharing
- overview_2_new_user_guide.md: add a Data Cubes entry to the dataset-type
  routing list

No API change; there is deliberately no cube dataset type to document.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015zxc8vHNQ6FPqsRJaT8tGk

Author: claude

docs/overview/overview_2_new_user_guide.md

@@ -52,6 +52,7 @@ interested in:
 
 - **CDD Imaging**: For image data from telescopes like Hubble and James Webb, go to [imaging/start_here.ipynb](https://github.com/PyAutoLabs/autolens_workspace/blob/main/notebooks/imaging/start_here.ipynb).
 - **Interferometer**: For radio / sub-mm interferometer from instruments like ALMA, go to [interferometer/start_here.ipynb](https://github.com/PyAutoLabs/autolens_workspace/blob/main/notebooks/interferometer/start_here.ipynb).
+- **Data Cubes**: For spectral-line data cubes (e.g. ALMA CO cubes), where many frequency channels are fitted simultaneously with a shared lens model, go to [interferometer/features/datacube/start_here.ipynb](https://github.com/PyAutoLabs/autolens_workspace/blob/main/notebooks/interferometer/features/datacube/start_here.ipynb).
 - **Point Sources**: For strongly lensed point sources (e.g. lensed quasars, supernovae), go to [point_source/start_here.ipynb](https://github.com/PyAutoLabs/autolens_workspace/blob/main/notebooks/point_source/start_here.ipynb).
 
 ## Google Colab

docs/overview/overview_3_features.md

@@ -13,6 +13,8 @@ links to the relevant workspace examples.
 
 **Interferometry**: Modeling of interferometer data (e.g. ALMA, LOFAR) directly in the uv-plane.
 
+**Data Cubes**: Modeling spectral-line data cubes (e.g. ALMA CO cubes), fitting every channel simultaneously with a shared lens model.
+
 **Multi Gaussian Expansion (MGE)**: Decomposing the lens galaxy into hundreds of Gaussians, for a clean lens subtraction.
 
 **Groups**: Modeling group-scale strong lenses with multiple lens galaxies and multiple source galaxies.
@@ -103,6 +105,22 @@ such as correlated noise. This uses the non-uniform fast fourier transform algor
 
 Checkout the `autolens_workspace/*/interferometer` package to get started.
 
+## Data Cubes
+
+Spectral-line observations produce a data cube: the same field observed in many adjacent frequency channels, so that
+the source's emission-line kinematics can be studied alongside the lens model.
+
+A cube is modeled as a list of `Interferometer` datasets, one per channel, combined into a single fit with
+`af.FactorGraphModel`. Every channel shares one lens mass model, while each channel reconstructs its own source, so
+the lens is constrained by the whole cube simultaneously rather than channel by channel. There is no separate cube
+dataset type to learn — the existing `Interferometer` and `AnalysisInterferometer` objects are reused throughout.
+
+Channel-invariant quantities (the lensing operator's curvature matrix) are computed once and shared across channels
+rather than rebuilt per channel, which is what makes fitting a many-channel cube tractable.
+
+Checkout the `autolens_workspace/*/interferometer/features/datacube` package to get started, which includes a
+`data_preparation` example covering the conversion from a CASA-style 4D FITS cube to the inputs the fit expects.
+
 ## Multi Gaussian Expansion (MGE)
 
 An MGE decomposes the light of a galaxy into tens or hundreds of two dimensional Gaussians: