Remove obsolete in-ShapePipe rho-stats / PSF-diagnostic plotting

[cailmdaley]

What

Removes the dead in-ShapePipe PSF-systematics plotting path (mean-shape focal-plane plots + rho statistics).

Martin confirmed on #657 (2026-04-22): "the rho stats within shapepipe are obsolete, and have been superseded by sp_validation / cosmo_val, which uses shear_psf_leakage." Per the stack division — ShapePipe produces catalogues, sp_validation validates — this path is deleted rather than migrated or preserved.

Changes

• Deleted modules/mccd_plots_runner.py and mccd_package/mccd_plot_utilities.py. Worth noting: despite the docstring's mention of "rho statistics", the code only ever computed mean shapes and ellipticity histograms — there is no treecorr/stile correlation-function code in it. The module is auto-discovered by filename glob, so deleting the file fully deregisters it.
• Deleted example/cfis/config_Pl_mccd.ini and config_valjoint_Pl_mccd.ini — these existed solely to run mccd_plots_runner. Rather than leave them as husks with an empty MODULE = line, they're removed whole.
• Edited the six remaining configs that referenced the runner (config_exp_mccd, config_MsPl_psfex, config_MsPl_mccd, config_MsPl_stars, cfis_simu/config_tile_Sx_exp_mccd, cfis/defunct/config_tile_Sx_exp_mccd): dropped mccd_plots_runner from the MODULE list and removed the orphaned [MCCD_PLOTS_RUNNER] section.
• Edited mccd_package/__init__.py (removed the MCCD_PLOTS_RUNNER docstring block and the mccd_plot_utilities __all__ entry) and the doc references in pipeline_tutorial.md / post_processing.md, pointing PSF diagnostics at sp_validation/cosmo_val.

Two deliberate scope decisions (please sanity-check)

1. stile was already gone. It is not referenced anywhere in src/, example/, docs/, pyproject.toml, or uv.lock — it was never a declared dependency and isn't imported. Nothing to remove; flagging so the absence isn't read as an oversight.

2. random_cat is deliberately kept. The cleanup brief raised an open question — is random_cat rho-stats-only? On inspection it is a general random-catalogue generator (random points within the survey mask, healpix output; config_Rc.ini) for clustering / LSS null tests — Landy-Szalay-style randoms, not PSF rho statistics. It's authored by Martin and is independent of the removed plotting path, so bundling its removal would have overreached what was flagged as obsolete. If you do want it gone too, happy to do it in a follow-up.

Verification

In-image pytest suite green against this branch — 250 passed (run via Apptainer sandbox from ghcr.io/cosmostat/shapepipe:develop on candide). Structural tier (config-parse, runner-metadata, imports) covers the dangling-config / dangling-import cases.

grep -rIn "mccd_plots_runner\|MCCD_PLOTS_RUNNER\|mccd_plot_utilities\|stile\|rho_stats" src/ example/ docs/  →  clean

🤖 Generated with Claude Code

— Claude on behalf of Cail

[cailmdaley]

Closing this — on review it over-reaches what was actually authorized, and the part that was authorized already shipped.

The rho-stats removal Martin signed off on (#657: "option 1, delete") landed in #715 (merged 2026-04-23), which surgically stripped the rho-stats code out of mccd_plot_utilities.py / mccd_plots_runner.py and dropped stile/treecorr. As of develop, those modules carry no rho/stile/treecorr code — they're purely plot_meanshapes() / mean_shapes_plot() (focal-plane ellipticities, sizes, residuals, 1D histograms).

So everything this PR removes is the meanshapes/ellipticity plotting, not rho-stats. And on #715 Martin was explicit about that path:

Let's keep meanshapes, this is very useful and can be run on merged star and PSF catalogues as well. Not sure whether this is what @fabianhervaspeters is using for the 2D plots in the catalogue paper.

Deleting it here would directly contradict that, and potentially break a figure path for the catalogue paper. Net, there's nothing left to salvage in this PR — closing rather than trimming.

Apologies for the noise, @martinkilbinger — the branch was cut against a stale assumption that mccd_plots_runner was still the rho-stats path.

— Claude on behalf of Cail