feat: Phase 1 — mirror JIT likelihood profiling scripts

[Jammy2211]

Summary

Phase 1 of the autolens_profiling z_feature roadmap — mirrors the JIT likelihood profiling scripts and their input datasets from autolens_workspace_developer/jax_profiling/jit/ into this repo at likelihood/ and dataset/. _developer remains the source of truth; nothing is moved or deleted upstream.

Closes #1.

What changed

• 9 likelihood profiling scripts + 1 __init__.py mirrored at likelihood/{imaging,interferometer,point_source,datacube}/. Filenames preserved verbatim. The step-by-step profile narrative each script prints is unchanged.
• 14 input dataset files mirrored at dataset/{imaging/hst, interferometer/sma, point_source/simple}/. Checksums identical to source.
• 5 READMEs authored: top-level likelihood/README.md (replacing the placeholder from Phase 0) plus one per subfolder. Each per-section README covers what the likelihood path profiles, scripts available, default dataset, and a headline run-time table placeholder for Phase 4's dashboard.
• .gitignore extended to cover regenerable dataset variants (dataset/{cluster,group,multi}/, */simple/ data files) and per-run profiling artifacts under results/likelihood/.

Path rewrites (mechanical, uniform across all 9 scripts)

Pattern	Before	After
Dataset path	Path("jax_profiling") / "dataset" / ...	Path("dataset") / ...
Workspace-root walk-up	_script_dir.parents[2]	_script_dir.parents[1]
Results path	_workspace_root / "jax_profiling" / "results" / "jit" / <type>	_workspace_root / "results" / "likelihood" / <type>
Docstring sibling refs	``jax_profiling/{jit,imaging,interferometer,…}/``	``likelihood/...`` and `dataset/...`

The artifact filename convention is preserved verbatim: <script>_likelihood_summary_<instrument>_v<al.__version__>.{json,png}.

Dataset regeneration is documented, not bundled

The new repo does not include dataset_setup/ regeneration scripts (out of scope for Phase 1). The if should_simulate(...) block at the top of each script is replaced with a clear FileNotFoundError that points readers back at the source-of-truth regen scripts in autolens_workspace_developer/jax_profiling/dataset_setup/. Default smoke runs use the pre-mirrored HST / SMA / simple datasets and never trigger this path.

Decisions captured

• Dataset folder is top-level dataset/, not likelihood/dataset/. Reason: Phase 2 (simulators) and Phase 3 (Nautilus) will consume the same input data and a shared folder avoids duplication.
• No CI in this PR. Phase 5 lands the GitHub Actions for lint + profile re-runs.

Test plan — smoke results (all 4 ran end-to-end on CPU)

Script	Artifacts at expected path?	Regression assertion
likelihood/imaging/mge.py	✓ mge_likelihood_summary_hst_v2026.5.14.2.{json,png}	⚠️ pre-existing drift (got 27542.08, expected 27379.39; same drift in _developer/jax_profiling/jit/imaging/mge.py — script logic identical, constant EXPECTED_LOG_LIKELIHOOD_HST unchanged)
likelihood/interferometer/mge.py	✓ mge_likelihood_summary_sma_v2026.5.14.2.{json,png}	✓ PASSED
likelihood/point_source/image_plane.py	✓ image_plane_summary_v2026.5.14.2.{json,png}	⚠️ pre-existing drift (got -362.21, expected 0.07; large drift, but constant EXPECTED_LOG_LIKELIHOOD_IMAGE_PLANE unchanged from _developer — same script logic)
likelihood/datacube/delaunay.py	✓ delaunay_likelihood_summary_sma_v2026.5.14.2.{json,png}	✓ PASSED (both eager and full-pipeline cube assertions)

Mirror-vs-source diff verified clean: only the documented path rewrites and the should_simulate-block replacement changed. The regression-constant drifts pre-date this PR and would manifest if the corresponding scripts in autolens_workspace_developer/jax_profiling/jit/ were rerun on PyAutoLens 2026.5.14.2. They are independent of the mirror and should be triaged in their own issue against _developer (or the constants refreshed and re-committed if the new values are correct).

Phase 0 bootstrap commit (for context): 0087d6a