config(env_vars): repair FAST_PLOTS coverage for visualization scripts

[Jammy2211]

Summary

Several env_vars.yaml override entries were missing PYAUTO_FAST_PLOTS from their unset: list, so the build server kept PYAUTO_FAST_PLOTS=1 set even for the visualization scripts that explicitly assert PNGs land on disk. PYAUTO_FAST_PLOTS=1 short-circuits subplot_save / save_figure in PyAutoArray — no PNG ever written, assertions fail with "fit.png was not produced" / "dataset.png missing" (triage Cluster A, plus the parked 40d-old imaging/visualization NEEDS_FIX entry).

This PR closes the coverage gap for every affected script and adds a new override for point_source/visualization.py (which had no entry at all).

Configs Changed

• config/build/env_vars.yaml:
  • Updated pattern-convention header comment to document .py-anchored patterns.
  • imaging/visualization.py: appended PYAUTO_FAST_PLOTS to unset: (was only unsetting PYAUTO_SMALL_DATASETS).
  • interferometer/visualization.py: appended PYAUTO_SMALL_DATASETS to unset: — script loads FITS at full res with an explicit (100, 100) shape_native mask, so the 15×15 cap mismatches even when shape_native is explicit.
  • imaging/visualization_jax: appended PYAUTO_FAST_PLOTS to unset:.
  • point_source/visualization_jax: appended PYAUTO_FAST_PLOTS to unset:.
  • Added new override for point_source/visualization.py → unset: [PYAUTO_FAST_PLOTS] (JSON dataset, no FITS, no mask — PYAUTO_FAST_PLOTS alone is sufficient).

Upstream PR

PyAutoBuild #91 — depends on its _pattern_matches fix to make the existing imaging/visualization.py and interferometer/visualization.py overrides fire (previously dead because the matcher stripped .py from the path before substring matching).

Test Plan

• End-to-end smoke under autobuild-resolved env (full build_env_for_script pipeline, with PyAutoBuild feat: autolens point_source JAX visualization scripts #91 applied): all 5 affected scripts pass — imaging 64.1s, interferometer 69.5s, point_source 20.0s, point_source_jax 41.6s, imaging_jax 123.8s.
• /smoke_test autolens_test runs cleanly (visualization scripts are commented out in smoke_tests.txt for an unrelated bypass-mode issue, so this PR is not directly exercised by smoke).

Next triage cluster regen should clear the autolens entries of Cluster A and the parked imaging/visualization NEEDS_FIX.

🤖 Generated with Claude Code