feat: build canonical 220-task core retrieval benchmark manifest

Adds the complete core benchmark pipeline: n_repos derivation from Dockerfile
topology, verifier-quality classification (core_ready/conditional/extension_only)
from ABC audit signals, 220-task manifest with suite allocation targets, power
validation (0.696 overall, need 283 for 0.80), and canonical vs extension policy.

New scripts: derive_n_repos.py, generate_verifier_labels.py, build_core_manifest.py,
validate_core_manifest.py. New configs: core_benchmark_manifest.json,
verifier_quality_scheme.json, verifier_quality_labels.json.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sjarmak

configs/core_benchmark_manifest.json

@@ -0,0 +1,88 @@
+{
+  "manifest_total": 220,
+  "target_total": 220,
+  "shortfall": 0,
+  "distribution": {
+    "loc_bands": {
+      "2M-8M": 101,
+      "400K-2M": 79,
+      "8M-40M": 32,
+      "<400K": 5,
+      ">40M": 3
+    },
+    "n_repos": {
+      "1": 111,
+      "2": 18,
+      "3": 31,
+      "4": 60
+    },
+    "single_repo": 111,
+    "multi_repo": 109,
+    "retrieval_sensitive": 200,
+    "control": 20,
+    "suites": {
+      "csb_org_compliance": 12,
+      "csb_org_crossorg": 10,
+      "csb_org_crossrepo": 8,
+      "csb_org_crossrepo_tracing": 20,
+      "csb_org_domain": 10,
+      "csb_org_incident": 22,
+      "csb_org_migration": 23,
+      "csb_org_onboarding": 16,
+      "csb_org_org": 10,
+      "csb_org_platform": 6,
+      "csb_org_security": 30,
+      "csb_sdlc_debug": 4,
+      "csb_sdlc_design": 4,
+      "csb_sdlc_document": 2,
+      "csb_sdlc_feature": 6,
+      "csb_sdlc_fix": 18,
+      "csb_sdlc_refactor": 4,
+      "csb_sdlc_secure": 6,
+      "csb_sdlc_test": 6,
+      "csb_sdlc_understand": 3
+    },
+    "verifier_quality": {
+      "conditional": 167,
+      "core_ready": 53
+    }
+  },
+  "power_analysis": {
+    "assumptions": {
+      "effect_size": 0.05,
+      "alpha": 0.05,
+      "sigma_overall": 0.3,
+      "sigma_loc_subgroup": 0.28,
+      "sigma_repo_subgroup": 0.32
+    },
+    "overall": {
+      "n": 220,
+      "power": 0.696
+    },
+    "loc_small_repos": {
+      "n": 84,
+      "power": 0.373
+    },
+    "loc_large_repos": {
+      "n": 136,
+      "power": 0.549
+    },
+    "single_repo": {
+      "n": 111,
+      "power": 0.377
+    },
+    "multi_repo": {
+      "n": 109,
+      "power": 0.371
+    },
+    "min_n_80pct_power": {
+      "overall": 283,
+      "subgroup_loc": 247,
+      "subgroup_repo": 322
+    }
+  },
+  "recommendations": [
+    "Overall power 0.696 < 0.80. Need 283 tasks minimum.",
+    "Single-repo subgroup power is low. Consider whether single-repo moderation is testable."
+  ]
+}