fix: repair three stale plot API calls in chapter 4 tutorial 3

[Jammy2211]

Summary

Fixes the workspace-smoke failure in the public chapter_4_pixelizations teaching notebook:

TypeError: plot_array() got an unexpected keyword argument 'mask'

Reported from PyAutoHeart workspace-smoke run 30858578587, job smoke / run_notebooks (3.12, howtogalaxy, chapter_4_pixelizations), 2026-08-03. That run installed released autolens 2026.7.29.2; the newer 2026.8.4.1 does not fix it — PyAutoGalaxy main is at the release commit bf91c570 and the wrapper still rejects mask.

The caller was wrong, not the signature, so no library change is involved.

Two further stale-API breaks were hidden behind the reported one, because execution stopped at the first. The tutorial is only green now because it was re-run after each fix:

Stale call	Why it broke	Fix
plot_array(..., mask=mask)	autogalaxy/util/plot_utils.py:124 has no mask param. The overlay is auto-derived from array.mask at autoarray/plot/array.py:128, but auto_mask_edge returns None for an unmasked array — and the call sat before apply_mask.	Move below apply_mask, drop the kwarg
aplt.subplot_image_and_mapper	Not re-exported by autogalaxy.plot; lives in autoarray.plot	Use aaplt. — the convention tutorial_2_mappers.py in this same chapter already uses
inversion.reconstruction_to_native	Attribute no longer exists	mapped_reconstructed_operated_data (the plottable Array2D)

Moving the plot rather than deleting the kwarg restores the mask boundary the author intended, measured on the installed stack:

unmasked   -> mask.is_all_false: True  | auto_mask_edge: None
after mask -> mask.is_all_false: False | auto_mask_edge: (156, 2)

The private-path import from autoarray.inversion.plot.inversion_plots import subplot_of_mapper is also replaced by the public aaplt export, so one import now covers both helpers.

A sweep of every .py and .ipynb across all ~25 workspace repos found the mask= kwarg at exactly one logical site — HowToLens and HowToFit are clean.

Scripts Changed

• scripts/chapter_4_pixelizations/tutorial_3_inversions.py — moved plot_array below apply_mask and dropped mask=; repointed subplot_image_and_mapper and subplot_of_mapper at aaplt; replaced inversion.reconstruction_to_native with mapped_reconstructed_operated_data
• notebooks/chapter_4_pixelizations/tutorial_3_inversions.ipynb — regenerated from the script via PyAutoHands (not hand-edited)

Test Plan

• All five chapter_4_pixelizations tutorials run green locally (the failing CI job covers the whole chapter, not just tutorial 3)
• Failure reproduced on unchanged input first, matching the CI trace exactly
• AST detector re-run on the post-fix tree: 0 out-of-signature plot_array kwargs (it correctly still flagged the notebook until it was regenerated)
• Regenerated notebook confirmed to carry every fix
• Smoke tests pass for all affected workspaces

Follow-ups filed (not in this PR)

• HowToLens ch4 t3 plots before apply_mask at lines 76/190, so its mask overlay never draws. Not a crash — a quality gap.
• The related aplt.Output drift in autocti_workspace_test (unverified — different library) and euclid_strong_lens_modeling_pipeline.

Generated by the PyAutoLabs agent workflow.