Refactor mju_combineSparse to eliminate temporary buffers.

Combine sparse vectors in-place by first counting total `nnz` and then working backwards from the end. This removes the need for temporary buffers in `mju_combineSparse` and its callers and speeds up the function by ~10%.

PiperOrigin-RevId: 902530210
Change-Id: I4f48c327103552ab968d3915399c6067367bec9f

src/engine/engine_support.c

@@ -420,17 +420,11 @@ void mj_mulM2(const mjModel* m, const mjData* d, mjtNum* res, const mjtNum* vec)
 void mj_addM(const mjModel* m, mjData* d, mjtNum* dst,
              int* rownnz, int* rowadr, int* colind) {
   int nv = m->nv;
+
   // sparse
   if (rownnz && rowadr && colind) {
-    mj_markStack(d);
-    mjtNum* buf_val = mjSTACKALLOC(d, nv, mjtNum);
-    int* buf_ind = mjSTACKALLOC(d, nv, int);
-
-    mju_addToMatSparse(dst, rownnz, rowadr, colind, nv,
-      d->M, m->M_rownnz, m->M_rowadr, m->M_colind,
-      buf_val, buf_ind);
-
-    mj_freeStack(d);
+    mju_addToMatSparse(dst, rownnz, rowadr, colind, nv, d->M,
+                       m->M_rownnz, m->M_rowadr, m->M_colind);
   }
 
   // dense