Cache group definitions that come from PDB files

[giacomofiorin]

Loading atom indices from a PDB file is relatively slow. Such a startup can be very tedious when using VMD on a large system, for example.

Although one can do the following:

colvar {
  function1 {
    atoms {
      name myatoms
      atomsFile file.pdb
    }
  }
  function2 {
    atoms { atomsOfGroup myatoms }
  }
}

it would not be too difficult to add a convenience function that automatically does that when it detects multiple reads of the same file.