vector vb = (inverse of matrix rm) * vector va
  (single precision)
  n.b.  The matrix must be unitary, as this routine assumes that
        the inverse and transpose are identical.
     rm       float[3][3]    matrix
     va       float[3]       vector
     vb       float[3]       result vector
  The same vector can be specified for both va and vb.
  P.T.Wallace   Starlink   31 October 1993