feat(eval): add field-level OCR accuracy harness + seed corpus

Adds a measurement gate for the artifact OCR parser (ADR-007), the prerequisite
for the layout-profile and preprocessing rework. runOcrEval feeds labeled OCR
text through the real parseArtifactCandidate and scores per-field / per-case
accuracy.

- src/eval/ocrEvalHarness.ts: pure metrics (per-field, critical-field, exact).
- src/eval/corpus/seedCorpus.ts: 23 cases transcribed from the verified parser
  test assertions; runs at 100%.
- src/eval/reviewSampleCorpus.ts: converts review samples into label *candidates*
  (never ground truth) so the review queue can grow the corpus.
- src/eval/ocrEval.test.ts + reviewSampleCorpus.test.ts: gate (must stay 1.0) and
  converter unit tests.
- npm run eval script; docs/ocr-eval.md; ADR-007/008/009.

Also records the agreed rework direction: C# input/capture sidecar (ADR-008) and
resolution-anchored layout profiles + OCR preprocessing (ADR-009).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/eval/ocrEval.test.ts

@@ -0,0 +1,44 @@
+import { describe, expect, it } from "vitest";
+import { formatReport, runOcrEval } from "./ocrEvalHarness";
+import { seedCorpus } from "./corpus/seedCorpus";
+
+// Regression gate: the seed corpus is verified ground truth, so the parser must
+// read every labeled field correctly. A drop here means an OCR/parser change
+// regressed a previously-correct read - look at the printed failures. If a
+// change intentionally alters a correct output, update the corpus label in the
+// same commit (the label is the source of truth, not the code).
+
+describe("OCR eval harness", () => {
+  const report = runOcrEval(seedCorpus);
+
+  it("prints the accuracy report", () => {
+    // Surfaced in test output for humans; not an assertion.
+    // eslint-disable-next-line no-console
+    console.log("\n" + formatReport(report) + "\n");
+    expect(report.totalCases).toBe(seedCorpus.length);
+  });
+
+  it("reads every labeled field on the seed corpus correctly", () => {
+    const failureSummary = report.failures
+      .map((failure) => {
+        const wrong = failure.fields
+          .filter((entry) => !entry.correct)
+          .map((entry) => `${entry.field}: expected "${entry.expected}" got "${entry.actual}"`)
+          .join("; ");
+        return `${failure.id} -> ${wrong}`;
+      })
+      .join("\n");
+
+    expect(report.fieldAccuracy, `field regressions:\n${failureSummary}`).toBe(1);
+    expect(report.criticalFieldAccuracy).toBe(1);
+    expect(report.exactRate).toBe(1);
+  });
+
+  it("labels every critical field at least once across the corpus", () => {
+    const criticalCoverage = report.perField.filter(
+      (entry) => ["name", "slot", "mainStat", "mainValue", "setName"].includes(entry.field) && entry.evaluated > 0,
+    );
+    // name is derived indirectly on most cases; slot/mainStat/mainValue/setName must be exercised.
+    expect(criticalCoverage.length).toBeGreaterThanOrEqual(4);
+  });
+});