Parallelize compute_residuals in IndexIVFPQ (#4654)

Summary:
Add OpenMP parallelization to compute_residuals function.
  This addresses the TODO comment about parallelization.

  Changes:
  - Add #pragma omp parallel for to residual computation loop
  - Use if(n > 1000) to avoid overhead for small batches
  - Each iteration is independent and thread-safe

Pull Request resolved: https://github.com/facebookresearch/faiss/pull/4654

Reviewed By: mnorris11

Differential Revision: D86234690

Pulled By: limqiying

fbshipit-source-id: 49c5ede2890d52a1d69705a72ada190df299e5d5

faiss/IndexIVFPQ.cpp

@@ -149,7 +149,8 @@ static std::unique_ptr<float[]> compute_residuals(
         const idx_t* list_nos) {
     size_t d = quantizer->d;
     std::unique_ptr<float[]> residuals(new float[n * d]);
-    // TODO: parallelize?
+    // Parallelize with OpenMP (each iteration is independent)
+#pragma omp parallel for if (n > 1000)
     for (size_t i = 0; i < n; i++) {
         if (list_nos[i] < 0)
             memset(residuals.get() + i * d, 0, sizeof(float) * d);