Model n-ary nodes better

[dlurton]

We currently model "n-ary" nodes as direct variants of the expr sum type:

(sum expr
    // Arithmetic operators
    (plus (* expr 2))
    (minus (* expr 2))
    (times (* expr 2))
    (divide (* expr 2))
    (modulo (* expr 2))
    (string_concat (* expr 2))

    // Logical operators
    (and (* expr 2))
    (or (* expr 2))

    // Comparison operators
    (eq (* expr 2))
    (ne (* expr 2))
    (gt (* expr 2))
    (gte (* expr 2))
    (lt (* expr 2))
    (lte (* expr 2))
    (in_container (* expr 2))
    ...)

However a better method of modeling them that allows for greater re-use of Kotlin code is to model the function as another sum variant:

(sum expr
    ...
    (nary nary_op (* expr 2))
    ...)

(sum nary_op)
    // Arithmetic operators
    (plus)
    (minus)
    (times)
    (divide)
    (modulo)
    (string_concat)

    // Logical operators
    (and)
    (or)

    // Comparison operators
    (eq)
    (ne)
    (gt)
    (gte)
    (lt)
    (lte)
    (in_container)
)