Add --verify-data CLI flag, Tensor.data property, parallel DataExecutor

- CLI: --verify-data flag enables Phase 2 data verification (ADR-0020)
- Tensor.data: returns actual numpy values (verify-data) or zeros placeholder
- Tensor.__repr__: shows value summary or data=N/A (placeholder)
- DataExecutor: ThreadPoolExecutor for same-timestamp parallel op execution
- BenchResult.engine: exposes op_log/memory_store for Phase 2 access

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

src/kernbench/sim_engine/data_executor.py

@@ -6,6 +6,7 @@ Same-timestamp independent ops can be batched for efficiency.
 """
 from __future__ import annotations
 
+from concurrent.futures import ThreadPoolExecutor
 from itertools import groupby
 from typing import Any
 
@@ -28,11 +29,18 @@ class DataExecutor:
         self.store = store
 
     def run(self) -> None:
-        """Execute all ops in op_log order, grouped by t_start."""
-        for _t, ops_iter in groupby(self._op_log, key=lambda r: r.t_start):
-            ops = list(ops_iter)
-            for op in ops:
-                self._execute_op(op)
+        """Execute all ops in op_log order, grouped by t_start.
+
+        Same-timestamp ops are independent and executed in parallel
+        via ThreadPoolExecutor (numpy releases the GIL for BLAS ops).
+        """
+        with ThreadPoolExecutor() as pool:
+            for _t, ops_iter in groupby(self._op_log, key=lambda r: r.t_start):
+                ops = list(ops_iter)
+                if len(ops) == 1:
+                    self._execute_op(ops[0])
+                else:
+                    list(pool.map(self._execute_op, ops))
 
     def _execute_op(self, op: OpRecord) -> None:
         if op.op_kind == "memory":