docs: comprehensive output folder layout in modeling tutorials and use search.paths in results/start_here

[Jammy2211]

Summary

Companion to PyAutoLabs/autolens_workspace#124 — the same two paired documentation improvements applied to the autogalaxy modeling tutorials:

1. Comprehensive output folder layout. Every modeling.py tutorial (imaging, interferometer, ellipse) replaces the brief flat-bullet __Output Folder__ description with a full directory-tree __Output Folder Layout__ block, adapted per data type and using galaxies.json rather than the autolens-specific tracer.json.

2. results/start_here.py no longer needs a <unique_hash> placeholder in code. The model-fit now runs once at the top of the file (mirroring _quick_fit.py / aggregator/samples.py), and the simple-loading section reaches results via result_path = search.paths.output_path. Users no longer have to hand-edit a 32-character identifier into the script before it runs.

The aggregator section keeps all its existing narrative. Its intro is rewritten to frame it as a first-class peer tool — generator-based, used by the csv_maker / png_maker / fits_maker workflow tools — rather than a "many fits" fallback.

Tracking issue: PyAutoLabs/autolens_workspace#124.

Scripts Changed

• scripts/imaging/modeling.py — replaced __Output Folder__ flat-bullet with full directory-tree __Output Folder Layout__ block (galaxies-not-tracer, no source-plane)
• scripts/interferometer/modeling.py — same treatment, adapted for visibility/uv-plane products (dirty_images.fits)
• scripts/ellipse/modeling.py — same treatment, adapted for ellipse-fit products (ellipse.json instead of galaxies.json)
• scripts/guides/results/start_here.py — restructured: model-fit now runs first; simple-loading uses search.paths.output_path instead of a hardcoded <unique_hash> Path; aggregator intro rewritten to frame it as a peer tool (generator-based, used by workflow tools); fixed pre-existing latent af.SamplesNest.from_csv → from_table API bug uncovered once the path resolved to a real folder

Test Plan

• python3 compile() syntax check passes for all 4 modified scripts
• scripts/guides/results/start_here.py runs end-to-end under PYAUTO_TEST_MODE=2 PYAUTO_SMALL_DATASETS=1 PYAUTO_FAST_PLOTS=1, reaching the final chi_squared / log_likelihood prints
• scripts/check_sizes.sh clean
• Smoke tests pass for the workspace as a whole

🤖 Generated with Claude Code