Fix extended stabilizer thread safety in apply_ops_parallel (#1993)

* Extended stabilizer simulator no longer shares RngEngine amongst states when ops are applied in parallel

* Added release note

* Fixed ugly cast

---------

Co-authored-by: Jun Doi <doichan@jp.ibm.com>

releasenotes/notes/fix_extstabilizer_thread_safety-c85e926c7ecb8dfb.yaml

@@ -0,0 +1,6 @@
+fixes:
+  - |
+    Extended stabilizer simulation was sharing a single copy of RngEngine amongst
+    parallelized states in ``ExtendedStabilizer::State::apply_ops_parallel``, 
+    leading to thread safety issue. Now, a new RngEngine is seeded for each parallel
+    state.

src/simulators/extended_stabilizer/extended_stabilizer_state.hpp

@@ -463,17 +463,24 @@ template <typename InputIterator>
 void State::apply_ops_parallel(InputIterator first, InputIterator last,
                                ExperimentResult &result, RngEngine &rng) {
   const int_t NUM_STATES = BaseState::qreg_.get_num_states();
+
+  std::vector<size_t> rng_seeds(NUM_STATES);
+  for (int_t i = 0; i < NUM_STATES; i++) {
+    rng_seeds[i] = rng.rand_int<size_t>(0, SIZE_MAX);
+  }
+
 #pragma omp parallel for if (BaseState::qreg_.check_omp_threshold() &&         \
                              BaseState::threads_ > 1)                          \
     num_threads(BaseState::threads_)
   for (int_t i = 0; i < NUM_STATES; i++) {
     if (!BaseState::qreg_.check_eps(i)) {
       continue;
     }
+    RngEngine local_rng(rng_seeds[i]);
     for (auto it = first; it != last; it++) {
       switch (it->type) {
       case Operations::OpType::gate:
-        apply_gate(*it, rng, i);
+        apply_gate(*it, local_rng, i);
         break;
       case Operations::OpType::barrier:
       case Operations::OpType::qerror_loc: