test(evals): retire every paid automatic retry

Paid evals never retry (approved 2026-09-29): delete SHORT_CASE_RETRY_FILES
and retriesWithinCaseCap, drop the retry fields from the registered wall rows
(walls now cover one run plus reserve), make retriesForFiles return 0, pass
--retry 0 explicitly, and drop --retry 1 from the package.json paid scripts.
Add the eval:pass-rates alias. Tests that pinned the old retry allowance are
updated as a policy change; review-finalization-budget now proves late-result
recording under the production zero-retry arguments.

test/carve-section-sharding.test.ts

@@ -22,7 +22,7 @@ describe('carved-skill cases each get a complete paid process budget', () => {
     expect(selectPaidTestFiles(files.map(file => 'test/' + file), 'periodic').selected).toHaveLength(files.length);
     expect(selectPaidTestFiles(files.map(file => 'test/' + file), 'gate').selected).toHaveLength(0);
   });
-  test('all configured retries plus teardown fit even with within-shard concurrency one', () => {
+  test('every case run plus teardown fits even with within-shard concurrency one', () => {
     for (const file of files) {
       const attempts = retriesForFiles(['test/' + file]) + 1;
       expect(CAPTURE_LONG_MS * attempts + 10_000).toBeLessThan(DEFAULT_SHARD_TIMEOUT_MS);