laplace is again proceeding to code generation

UFL did not like  my way of replacing some args with 1 as it
destroyed the shape information. It now remains in the expression,
but triggers code generation separately.

and remove old implementation