Fixed incorrect array size in diagMatSym

According to dsyevr documentation:
- the matrix being diagonalized is overwritten.
  this is not happning in practice, but I added a temporary matrix
- the vector of eigenvalues should hold N elements even if we
  only compute M<N elements.

The latter point is important when we diagonalize a 4x4 matrix and
only need the lowest eigenvalue (as in RMSD calculations).
Notice that in practice the next element of the array
is only accessed if there are two numerically identical eigenvalues,
which happens when atoms are aligned on a straight line
(unlikely)

src/tools/Tensor.h

@@ -535,6 +535,10 @@ void diagMatSym(const TensorGeneric<n,n>&mat,VectorGeneric<m>&evals,TensorGeneri
   std::array<int,10*n> iwork;
   std::array<double,(6+bs)*n> work;
   std::array<int,2*m> isup;
+  // documentation says that "matrix is destroyed" !!!
+  auto mat_copy=mat;
+  // documentation says this is size n (even if m<n)
+  std::array<double,n> evals_tmp;
   int nn=n;              // dimension of matrix
   double vl=0.0, vu=1.0; // ranges - not used
   int one=1,mm=m;        // minimum and maximum index
@@ -543,13 +547,14 @@ void diagMatSym(const TensorGeneric<n,n>&mat,VectorGeneric<m>&evals,TensorGeneri
   int info=0;            // result
   int liwork=iwork.size();
   int lwork=work.size();
-  TensorGenericAux::local_dsyevr("V", (n==m?"A":"I"), "U", &nn, const_cast<double*>(&mat[0][0]), &nn, &vl, &vu, &one, &mm,
-                                 &abstol, &mout, &evals[0], &evec[0][0], &nn,
+  TensorGenericAux::local_dsyevr("V", (n==m?"A":"I"), "U", &nn, const_cast<double*>(&mat_copy[0][0]), &nn, &vl, &vu, &one, &mm,
+                                 &abstol, &mout, &evals_tmp[0], &evec[0][0], &nn,
                                  isup.data(), work.data(), &lwork, iwork.data(), &liwork, &info);
   if(info!=0) plumed_error()<<"Error diagonalizing matrix\n"
                               <<"Matrix:\n"<<mat<<"\n"
                               <<"Info: "<<info<<"\n";
   plumed_assert(mout==m);
+  for(unsigned i=0; i<m; i++) evals[i]=evals_tmp[i];
   // This changes eigenvectors so that the first non-null element
   // of each of them is positive
   // We can do it because the phase is arbitrary, and helps making