Add tuple field access and tuple-shaped split results

[someone235]

This adds tuple field access for tuple-returning function calls, so callers can read individual return values directly with .0, .1, .2, and so on.

It also updates split(int) to behave like a tuple-returning built-in: its two results are accessed with .0 and .1, or destructured with parenthesized tuple binding syntax. As part of making tuple returns
explicit, parenthesized single return types like : (int) are now treated as one-element tuples, not plain scalar returns.

Language Behavior Changes

• Tuple-returning function calls can be accessed with numeric tuple fields.

function f() : (int, int, int, int) {
    return (2, 3, 4, 5);
}

entrypoint function main() {
    int x = f().2;
    require(f().3 == 5);
}

• One-element tuple returns use .0.

function f() : (int) {
    return (7);
}

entrypoint function main() {
    require(f().0 == 7);
}

• Plain scalar function returns remain distinct from tuple returns.

function scalar() : int {
    return 7;
}

function wrapped() : (int) {
    return (7);
}

• Tuple-returning functions cannot be used directly as scalar expressions.

function f() : (int) {
    return (7);
}

entrypoint function main() {
    require(f() == 7); // invalid; use f().0
}

• split(int) now uses tuple-style access.

byte[] left = data.split(4).0;
byte[] right = data.split(4).1;

• Split destructuring now uses parenthesized tuple binding syntax.

(byte[16] left, byte[16] right) = data.split(16);

• The old split index form is no longer the canonical syntax.

byte[] left = data.split(4)[0]; // use data.split(4).0