fix: change the setting of force in group_force_object to accumulation

This should better match what the existing apply_force does.

src/colvaratoms.cpp

@@ -1482,7 +1482,7 @@ void cvm::atom_group::apply_force(cvm::rvector const &force)
 
   auto ag_force = get_group_force_object();
   for (size_t i = 0; i < size(); ++i) {
-    ag_force.set_atom_force(i, atoms[i].mass / total_mass * force);
+    ag_force.add_atom_force(i, atoms[i].mass / total_mass * force);
   }
 }
 
@@ -1515,9 +1515,9 @@ cvm::atom_group::group_force_object::~group_force_object() {
   }
 }
 
-void cvm::atom_group::group_force_object::set_atom_force(size_t i, const cvm::rvector& force) {
+void cvm::atom_group::group_force_object::add_atom_force(size_t i, const cvm::rvector& force) {
   if (m_has_fitting_force) {
-    m_ag->group_forces[i] = force;
+    m_ag->group_forces[i] += force;
   } else {
     // Apply the force directly if we don't use fitting
     (*m_ag)[i].apply_force(force);

src/colvaratoms.h

@@ -266,7 +266,7 @@ class colvarmodule::atom_group
   public:
     group_force_object(cvm::atom_group* ag);
     ~group_force_object();
-    void set_atom_force(size_t i, const cvm::rvector& force);
+    void add_atom_force(size_t i, const cvm::rvector& force);
   private:
     cvm::atom_group* m_ag;
     cvm::atom_group* m_group_for_fit;

src/colvarcomp_distances.cpp

@@ -1408,7 +1408,7 @@ void colvar::cartesian::apply_force(colvarvalue const &force)
       for (j = 0; j < dim; j++) {
         f[axes[j]] = force.vector1d_value[dim*ia + j];
       }
-      ag_force.set_atom_force(ia, f);
+      ag_force.add_atom_force(ia, f);
     }
   }
 }

src/colvarcomp_rotations.cpp

@@ -144,7 +144,7 @@ void colvar::orientation::apply_force(colvarvalue const &force)
                         FQ[1] * dq0_2[1] +
                         FQ[2] * dq0_2[2] +
                         FQ[3] * dq0_2[3];
-      ag_force.set_atom_force(ia, f_ia);
+      ag_force.add_atom_force(ia, f_ia);
     }
   }
 }