May need some special treatment for the evaluation function 'accuracy'

[Zhang-Liao]

The definition is

evalfn(accuracy,[P,N|_],Val):-
	(P = -inf -> Val is 0.5;
		Val is P / (P + N)), !.

It causes an error when P + N = 0. It may happen when users disable the bottom clause and define their own refinement operators. It is possible that generating a clause that does not satisfy any examples.

Perhaps return accuracy 0 when P + N = 0?

[friguzzi]

I made it return 0.5 as for the other default case.
Fixed in 90b7837