Fixing bug where a term was tested for type using unification

[cmungall]

[samer--]

Hi Chris, can you explain a bit why you wanted to make this change? Unless there is corresponding bug somewhere else, this bit looks right as it is, because it's comparing the value to be type checked against a nullary constructor. For example, a maybe type like in Haskell would be declared as :- type maybe(A) ---> nothing; just(A). The presence of the nullary constructor nothing means that the term nothing has type maybe(A) for any type A. In such a case, user_type_constructor(maybe(A), Cons) will succeed with Const=nothing, and hence a term Term being type checked will have type maybe(A) if Term unifies with nothing, not when error:has_type(nothing, nothing), which should never succeed, since nothing is not a type.