Paper figures for the DESC internal review, plus example fixes

[eduardojsbarroso]

Summary

Updates the cosmoDC2 redMaPPer and capish example notebooks so that the paper figures answer the figure comments of the DESC internal review, and fixes a few example scripts. The notebooks were rerun and the paper figures regenerated.

Paper figures

• Figures are drawn at their printed width (7.1 in or 3.4 in) with darker grid lines and larger legends.
• Shared style helpers in clpipe/utils/cosmosis_mcmc_plots.py: page widths, one color and marker per richness bin (Okabe-Ito based, checked for color-blind readers), richness_label, add_grid, use_paper_style and thicken_legend_lines. Corner plots get larger legends.
• cosmosis_mcmc_plots now keeps the notebook backend when it imports getdist. Older getdist versions switch matplotlib to Agg on import, which stopped figures from showing inline.

Paper figure	Notebook	Change
Fig. 3 precision_plot.pdf	crow_clcosmo_benchmark	drop leftover "CROW grid" legend, richness colors and markers
Fig. 4 counts_shear_capish_z0.5_0.6.pdf	capish analysis_summary	counts prediction as steps per bin, error bars, R = 1 Mpc tick, legend inside the lensing panel
Figs. 6 and 7	prediction_versus_data	counts predictions as steps, richness colors and markers, R = 1 Mpc tick
Figs. 8 and 14	plot_samples	larger legends, Fig. 14 labels "This work (CLPipe)" and "Payerne et al. (2025)"
Fig. 9 summary_cosmo.pdf	summary_plot_best_fits	larger value labels
Fig. 10	mor_fiducial	no redshift colors, dashed models with one "Model" legend entry
Fig. 11 miscentering.pdf	miscentering_fiducial	log scale for the offset histogram
Fig. 12	shear_profiles_different_cosmologies	signed fractional difference with a ±5% band
Fig. 13 10mpc_clpipeline_vs_clcosmo.pdf	data_clcosmosim_txpipe	full page width, new labels, R [Mpc] axis

Other changes

• fiducial_cosmology_analysis: the cosmo_0.22_0.7, cosmo_0.3_0.9 and cosmo_0.4_0.8 launch scripts run cosmosis with MPI like the other runs, and the missing cosmo_0.22_0.9 launch script is added.
• capish analysis_summary: 15% burn-in and mu_m = 2.3/ln(10). run_in2p3_both_mean/TJPCov.yml reads the mean SACC from data/.
• plot_samples: the cosmoDC2 contour uses a text label instead of the legend.