Add chunked parallel mapping function

  Should be more efficient than the default parallel mapping function,
  if no sideeffects are caused by the mapped function.