fix: restore unconditional dataset.fits output for aggregator (save_attributes)

[Jammy2211]

Summary

The aggregator result guides (guides/results/aggregator/models.py) fail in both autogalaxy_workspace and autolens_workspace with:

PyAutoGalaxy/autogalaxy/aggregator/agg_util.py:101
  header = aa.Header(header_sci_obj=fit.value(name=name)[0].header)
TypeError: 'NoneType' object is not subscriptable

Surfaced by the 2026-07-08 full workspace-validation sweep (5 failed / 613 passed).

Root cause

fit.value(name="dataset") returns None: the fit no longer persists the dataset.fits the aggregator loaders require.

Pre-overhaul, AnalysisDataset.save_attributes wrote dataset.fits to the files/ folder unconditionally (before every fit). The "Visualization overhaul: fits API" commit (081bbd84, via 825544bf "move save attributes stuff to plotter interface") removed that write and left only the plotter-interface copy in image/, which is gated on the fits_dataset visualization setting. Any fit that does not emit that visualization output (test-mode / minimal searches / older runs) has no dataset.fits, so ImagingAgg / InterferometerAgg / agg_util.mask_header_from break.

fit.value scans both files/ and image/, so where visualization ran the file is found — the failure is on fits where it did not.

PyAutoLens's analysis never had a save_attributes producer at all (it relied entirely on the visualization copy), so the same gap affects autolens_workspace.

Fix (restore the producer; both libraries)

Re-add an unconditional paths.save_fits(name="dataset", ...) to save_attributes, honouring the documented HDU contract (mask/data/noise_map/psf/over_sample_size_* for imaging; mask/data/noise_map/uv_wavelengths for interferometer), using the current plotter accessors (.native_for_fits, .header_dict):

• PyAutoGalaxy AnalysisImaging + AnalysisInterferometer.save_attributes
• PyAutoLens AnalysisImaging (new override) + AnalysisInterferometer.save_attributes

No consumer-side None-guard (fix the producer). The plotter's image/dataset.fits write is unchanged.

Verification

• autogalaxy and autolens guides/results/aggregator/models.py both run to completion (exit 0), TypeError gone.
• New regression tests assert save_attributes writes files/dataset.fits with [MASK, DATA, NOISE_MAP, PSF].
• Existing analysis suites pass (PyAutoGalaxy 5, PyAutoLens 10).

[Jammy2211]

Addressed by two library-first PRs (both pending-release):

• PyAutoGalaxy fix: restore unconditional dataset.fits output for aggregator (save_attributes) #479 (producer fix + regression test)
• PyAutoLens #574 (companion fix + regression test)

Verified end-to-end: both guides/results/aggregator/models.py scripts (autogalaxy + autolens) now run to completion. Merge #479 first, then #574.