Update anova_recompute.py default output path to anova.json

Add anova.json: two-way Type II ANOVA on delta_l_ortho with model x complexity-level factors (paper §4.3)

Add wilcoxon.json: per-pair Wilcoxon signed-rank tests with Bonferroni correction (paper Table 3 caption: 153/180 pairs significant at p<0.01)

Remove significance.json (split into wilcoxon.json + anova.json for clarity)

Update anova_recompute.py to write significance.json (no version suffix)

Remove redundant neg_prompt_ablation_v2.json (paper-version content now lives in evaluation/neg_prompt_ablation.json)

Remove redundant significance_v2.json (paper-version content now lives in evaluation/significance.json)

Update neg_prompt_ablation.json to V5-mask version (used in paper)

Update significance.json to Bonferroni-corrected version (used in paper)

Update anova_recompute.py with --input/--output CLI flags

Add evaluation/anova_recompute.py

Add evaluation/significance_v2.json

Add evaluation/significance.json

Add evaluation/fid_lpips.json

Add evaluation/seed_ablation.json

Add evaluation/neg_prompt_ablation_v2.json

Add evaluation/neg_prompt_ablation.json

Add evaluation/n3_lvp_improved.json

Add evaluation/n1_resolution_ablation.json

Add evaluation/mde_ablation_summary.json

Add evaluation/mde_ablation_sampled_views.json

Add evaluation/mde_ablation_results.json

Add evaluation/error_decomposition.json

Add evaluation/igf_results.json

Add evaluation/igf_summary.json