[Relay] Move prelude to text format

[weberlo]

This PR is a follow-up to #3863, which replaces the hand-constructed ASTs in the current prelude with equivalent definitions in the text format.

One question I have about syntax is how we should handle zero-arity constructors. Most functional languages desugar, for example, Nil into Nil(), and I've implemented that functionality here. Let me know if that's not what we want.

CC @MarisaKirisame @slyubomirsky @jroesch @junrushao1994

[slyubomirsky]

Good idea, though it probably merits wider discussion

[jroesch]

cc @zhiics and @wweic

[zhiics]

Will do over the weekend.

[zhiics]

Isn't this function the same as before?

[weberlo]

yeah. I just moved it next to Add for readability. I can move it back if you want.

[wweic]

LGTM. I'll approve once CI is fixed.

[weberlo]

The CI failure was because the VM compiler can't compile programs which make use of first-class constructors. For example,

/*
 * Concatenates two lists.
 */
def @concat[A](%xs: List[A], %ys: List[A]) -> List[A] {
  @foldr(Cons, %ys, %xs)
}

needs to be rewritten as

/*
 * Concatenates two lists.
 */
def @concat[A](%xs: List[A], %ys: List[A]) -> List[A] {
  let %updater = fn(%x: A, %xss: List[A]) -> List[A] {
    Cons(%x, %xss)
  };
  @foldr(%updater, %ys, %xs)
}

, in order to successfully compile.

@jroesch We do want first-class constructors, right? If so, I'll submit a follow-up PR that adds that functionality.

[weberlo]

@wweic CI is green

[jroesch]

The CI failure was because the VM compiler can't compile programs which make use of first-class constructors. For example,

/*
 * Concatenates two lists.
 */
def @concat[A](%xs: List[A], %ys: List[A]) -> List[A] {
  @foldr(Cons, %ys, %xs)
}

needs to be rewritten as

/*
 * Concatenates two lists.
 */
def @concat[A](%xs: List[A], %ys: List[A]) -> List[A] {
  let %updater = fn(%x: A, %xss: List[A]) -> List[A] {
    Cons(%x, %xss)
  };
  @foldr(%updater, %ys, %xs)
}

, in order to successfully compile.

@jroesch We do want first-class constructors, right? If so, I'll submit a follow-up PR that adds that functionality.

We should handle this with a full eta-expansion pass, I will try to find some resources on performing eta-expansion for these types of purposes.