feat: Phase 2 — mirror simulator profiling scripts

[Jammy2211]

Summary

Phase 2 of the autolens_profiling z_feature roadmap — mirrors the 6 simulator-profiling scripts from autolens_workspace_developer/jax_profiling/simulators/ (~2,040 LOC across imaging, interferometer, point_source, cluster, group, multi) into the new repo at simulators/. _developer stays the source of truth — the originals do not move.

Closes #4.

What changed

• 6 simulator scripts mirrored from clean origin/main of _developer/jax_profiling/simulators/ → autolens_profiling/simulators/.
• 4 path-rewrite patterns applied uniformly across all 6:
  • _workspace_root / \"jax_profiling\" / \"dataset\" → _workspace_root / \"dataset\"
  • _workspace_root / \"jax_profiling\" / \"results\" / \"simulators\" → _workspace_root / \"results\" / \"simulators\"
  • _script_dir.parents[1] → _script_dir.parents[0] (scripts are one level shallower in the new layout — simulators/<name>.py vs jax_profiling/simulators/<name>.py)
  • Docstring example commands: python jax_profiling/simulators/<name>.py → python simulators/<name>.py
• simulators/README.md replaces the Phase 0 placeholder. Covers: what each simulator profiles, the dataset_name collision warning for point_source.py, versioned artifact convention, headline run-times table placeholder for Phase 4's dashboard, and a per-script JIT-readiness audit.

Artifact filename convention is unchanged: results/simulators/<script>_summary_v<al.__version__>.{json,png}.

Test plan

• py_compile PASSES for all 6 mirrored scripts.
• grep -r jax_profiling returns no residual references in the mirrored scripts.
• dataset/<type>/<name>/ write target verified to resolve correctly with _script_dir.parents[0] (one level shallower than Phase 1's parents[1]).
• .gitignore from Phase 1 already covers the regenerable dataset variants (dataset/{cluster,group,multi}/, */simple/), so smoke runs that use default dataset_name = \"simple\" won't leak binary outputs.
• Full runtime smoke skipped: running simulators/point_source.py at default dataset_name=\"simple\" would overwrite Phase 1's tracked dataset/point_source/simple/{point_dataset_positions_only.json,tracer.json} (corrupting the likelihood inputs that Phase 1 specifically locked in). Phase 5 (issue #7) will land the AUTOLENS_PROFILING_SMOKE=1 env-var short-circuit that makes future smokes cheap and side-effect-free. Until then, smoke manually by passing a non-conflicting dataset_name (e.g. \"smoke\").

F1 lesson applied

Copies came from the worktree's clean origin/main of _developer — not the canonical (which is dirty with ~36 modified files, the same condition that motivated F1's re-mirror PR #3 for Phase 1).

Notes for Phase 4

• The simulator-side point_source.py exercises a JIT'd solver.solve and asserts eager ≡ JIT. This is the same assertion that drifted upstream and is tracked under PyAutoLens#514. Phase 4's dashboard cell for point_source/simulator may show ERR or pending upstream until that resolves — same handling as the likelihood-side point_source drift.

Refs

• Phase 0 (bootstrap, shipped): commit 0087d6a.
• Phase 1 (likelihood mirror, shipped): feat: Phase 1 — mirror JIT likelihood profiling scripts #2; F1 re-mirror fix: re-mirror likelihood scripts and datasets from clean origin/main of _developer #3.
• Phase 3 (Nautilus mirror, shipped): feat: Phase 3 — mirror Nautilus profiling scripts #8.
• Phase 4 (dashboard, queued): feat: Phase 4 — instrument-framed README dashboard #6 — will read this PR's simulator results.
• Phase 5 (CI, queued): feat: Phase 5 — GitHub Actions for lint + profile re-run #7 — provides the smoke short-circuit referenced above.
• F1 triage (shipped): _developer#67 + 3 upstream issues.

🤖 Generated with Claude Code