Harden paper plotting workflow and expose production coverage blocker

[Waxpardo]

Status

Draft / blocked: do not merge as a final-figure PR yet.

The plotting implementation and reduced smoke validation are working, but the exhaustive real-input audit proves that the current ten disjoint Nikhef subsamples cannot support the full paper observable set with N=10 finite yields.

What this changes

• keeps Paul Veen's grouped-trigger, THnSparse, multi-tune, mini/global-canvas architecture;
• makes both configs checkout-relative and schema-consistent;
• applies the documented ten-subsample SEM prescription and strict finite/nonzero checks in all four yield/ratio drawing paths;
• centralizes tune styling through TunePlotStyle.h and styles tune ratios by numerator tune;
• initializes/checks optional mini pads;
• adds input/provenance and exhaustive subsample-coverage validators;
• makes smoke a real-error validation of the universally supported 1-10% class;
• removes 78 stale tracked generated plot artifacts and keeps future plot output ignored;
• documents paper-figure provenance and the production blocker.

Validation

• ROOT object/inventory/manifests/union validation: passed.
• 56 central files per tune; 10 disjoint subsamples per tune; 56 files per subsample; job IDs 0-99 exactly once.
• 72 representative central-vs-subsample entry/integral comparisons: passed.
• Verbose reduced smoke: 30/30 statistics records have n=10, finite positive stdError, no missing inputs/objects, no NaN/inf, no zero normalization, no 1e-10 placeholders.
• Full coverage audit: failed as intended with 610 incomplete cases (540 beauty, 70 charm). Only 468/1152 logged yield/ratio statistics have n=10.

Blocking production result

Across every tune and associate in each trigger group:

• beauty B+ supports only integrated and 1-10%;
• beauty Lambda-b-bar supports no common bin;
• charm D+ supports integrated through 1-10%, but not 0-1%;
• charm Lambda-c supports integrated, 30-40%, 20-30%, and 1-10%.

The full target stays strict and fails rather than publishing incomplete, zero, or placeholder uncertainties. No smoke output was copied into the paper.

See PlottingScripts/validation/FINAL_PLOTTING_HANDOFF.md for commands, evidence, visual-QA findings, and the required next production step.