feat: Phase 4 — instrument-framed README dashboard

[Jammy2211]

Summary

Phase 4 of the autolens_profiling z_feature roadmap — builds the public-facing instrument-framed dashboard, the user's most explicitly-requested deliverable: "Framing it in terms of Astronomy instruments and telescopes, instead of just 'number of pixels' or other such metrics is way more intuitive."

Closes #6.

What lands in this PR

scripts/build_readme.py — idempotent table generator

• Scans results/**/*_summary_v<version>.json.
• Parses filenames into (section, sub-folder, script, instrument, version).
• Picks the latest version per axis via PEP 440-ish dotted-version sort.
• Generates a markdown table per known sentinel region and replaces content between matching <!-- BEGIN auto-table:<name> --> / <!-- END auto-table:<name> -->.
• --check mode for CI: exits non-zero if running would change any file (so Phase 5 can gate on it).

7 sentinel region types are wired up today:

Region	Target README
headline	top-level README.md
likelihood-imaging	likelihood/imaging/README.md
likelihood-interferometer	likelihood/interferometer/README.md
likelihood-point_source	likelihood/point_source/README.md
likelihood-datacube	likelihood/datacube/README.md
simulators	simulators/README.md
searches-nautilus	searches/nautilus/README.md

Sentinel-tagged auto-table regions added to 7 target READMEs

Hand-written prose around each region is preserved verbatim; only the body between BEGIN and END is rewritten. The placeholder "populated by Phase 4" tables from Phases 1-3 are gone, replaced by sentinel blocks + a one-line caption explaining how to regenerate.

Top-level README updated

• New "Latest run-times" section anchored on the headline auto-table.
• Roadmap table refreshed to show Phases 0-4 shipped (was Phase 0 only at bootstrap time).
• New "Future enhancements" subsection documenting six deferred extras: regression-watch indicator, per-axis version-history PNGs, plotly interactive timeline, flamegraph captures, hardware-tier columns, archive policy.

Today's output

No result artifacts are committed under results/ (they're gitignored per Phase 1's design — the artifact storage policy waits for Phase 5's CI workflow to lock in a commit cadence). Every auto-table currently renders a "No data yet — run X to populate" placeholder. As soon as the user or Phase 5's profile.yml workflow produces and commits real artifacts, re-running scripts/build_readme.py refreshes every table in place.

Design decisions resolved

• CPU/laptop-GPU/HPC-GPU split: single column for now (CPU only — every artifact today is implicitly CPU). Hardware-tier columns added as a "Future enhancements" entry; renderer change is small once artifacts gain a hardware label (filename suffix or JSON \"hardware\" field).
• Versioning policy: keep all versions in results/, sort by PEP 440-ish dotted version, render the latest. Archiving older versions to results/archive/ is a "Future enhancements" entry.
• "Cool extras": deferred all six candidates to "Future enhancements" rather than landing extras in this PR. Reasons: no real artifacts exist to feed them, the dashboard infrastructure is more valuable to ship first, and each extra is independently scoped under its own follow-up.

Test plan

• py_compile scripts/build_readme.py PASSES.
• python scripts/build_readme.py runs cleanly against the empty results/ tree and writes "no data yet" placeholders into every sentinel region.
• python scripts/build_readme.py --check exits 0 immediately after the previous run — idempotence verified.
• Surrounding hand-written prose in every target README is unchanged (verified by diff inspection — only content between sentinels was rewritten).
• Unknown sentinel names (none today) would be left intact with a stderr warning rather than erroring out — robust against future README additions.

Refs

• Phase 1 (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 2 (shipped): feat: Phase 2 — mirror simulator profiling scripts #9.
• Phase 3 (shipped): feat: Phase 3 — mirror Nautilus profiling scripts #8.
• Phase 5 (queued): feat: Phase 5 — GitHub Actions for lint + profile re-run #7 — will call python scripts/build_readme.py --check in lint.yml and python scripts/build_readme.py && git commit in profile.yml after profile-rerun.
• F1 upstream issues: PyAutoLens#514, _developer#68, _developer#69 — the dashboard's likelihood-point_source cells will likely show — until #514 resolves.

🤖 Generated with Claude Code