Distinguish failing vs stale review verdicts in check summary output

[Copilot]

npx rfc2119 check correctly emitted per-verdict diagnostics for failing reviews, but the final summary bucketed those failures as stale review(s), obscuring the required next action. This change separates failing verdicts from stale/missing verdicts in the summary line.

• Summary classification update (src/cli.ts)

  • Replaced summary use of report.staleReviews.length with explicit rule-based counts from report.violations:
    • REQ-003.2.4 → failing review(s)
    • REQ-003.3.1 → stale review(s)
  • Keeps total violation/uncovered counts unchanged while making review status actionable.

• Focused CLI behavior coverage (tests/cli.test.ts)

  • Added an end-to-end case that records a fail verdict and asserts summary output reports:
    • 1 failing review(s), 0 stale review(s)

const failingReviewCount = report.violations.filter((v) => v.rule === "REQ-003.2.4").length;
const staleReviewCount = report.violations.filter((v) => v.rule === "REQ-003.3.1").length;

• Fixes check summary labels failing review verdicts as stale reviews #6

[sansari]

@copilot instead of check: FAIL — 29 violation(s), 23 uncovered, 6 review failure(s), 0 stale review(s) make it like this:

check: FAILED w/ 29 violation(s) — 23 uncovered, 6 review failure(s), 0 stale review(s)