feat(eval): print per-case + total wall-clock time

Useful as prompts grow longer and we benchmark alternate models — bare
PASS/FAIL hides regressions like a prompt that doubles latency without
changing correctness.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

eval/run_eval.py

@@ -12,6 +12,7 @@ Run from the repo root:
 from __future__ import annotations
 
 import sys
+import time
 from pathlib import Path
 
 import yaml
@@ -45,24 +46,30 @@ def main() -> int:
         return 2
 
     failures = 0
+    total_start = time.perf_counter()
     for case_path in cases:
         case = yaml.safe_load(case_path.read_text())
         name = case["name"]
+        case_start = time.perf_counter()
         result = classify("db_kind", case["input"], DbKindResult)
+        elapsed = time.perf_counter() - case_start
         actual = result.model_dump()
         diffs = _diff_fields(case["expected"], actual)
+        status = "PASS" if not diffs else "FAIL"
+        print(f"{status} {name} ({elapsed:.2f}s)")
         if diffs:
             failures += 1
-            print(f"FAIL {name}")
             for line in diffs:
                 print(line)
-        else:
-            print(f"PASS {name}")
+    total_elapsed = time.perf_counter() - total_start
 
     if failures:
-        print(f"\n{failures}/{len(cases)} cases failed.", file=sys.stderr)
+        print(
+            f"\n{failures}/{len(cases)} cases failed. Total {total_elapsed:.2f}s.",
+            file=sys.stderr,
+        )
         return 1
-    print(f"\nAll {len(cases)} cases passed.")
+    print(f"\nAll {len(cases)} cases passed. Total {total_elapsed:.2f}s.")
     return 0