type system: formalize 'string | null' (nullable string) with proper narrowing — unblocks clean FFI

[cs01]

Summary

ChadScript's type system supports union types (per chadscript.d.ts: "Union Types — string | number support"), but string | null (nullable string) is not formalized as a first-class pattern with proper narrowing. This bites FFI and any code that needs "maybe-a-value" semantics.

Why it matters

Every library that calls C code hits this. C convention: return NULL to mean "no value" / "error". Currently chad's only options are:

1. Coerce NULL → "" (lossy — can't distinguish "unset" from "empty string"; see ffi: NULL char* does not reliably round-trip to empty string in TS #591)
2. Add a separate probe function (isOpen(), hasValue()) on every wrapper

Neither is ergonomic for user code. string | null with TypeScript-style narrowing (if (x !== null) { x.foo() }) is the right shape.

What "support" means concretely

1. string | null as a valid type annotation that typechecks and compiles
2. Null-literal assignment: const x: string | null = null;
3. Narrowing: after if (x !== null), x has type string and methods work
4. FFI bridge returns with i8* declared as string | null map NULL → null, non-null → string
5. const empty: string = ""; const missing: string | null = null; empty === missing → false

Why it's probably already near-implemented

Chad has union types, narrowing patterns for other unions, and NaN/tagged sentinel infrastructure for JS-compatible null/undefined. The pieces exist. The question is whether string | null specifically round-trips cleanly through the type system and codegen.

Impact

Unblocks:

• Cleaner FFI surface for every C library wrapper
• Optional return values from user code (function lookup(key: string): Record | null)
• Eliminates the isOpen() / hasValue() probe pattern from lib/net.ts, future Postgres driver, etc.
• Better TypeScript compatibility (nullable types are ubiquitous in real TS code)

Related

• ffi: NULL char* does not reliably round-trip to empty string in TS #591 — NULL char* FFI (the current workaround is coerce-to-empty-string; fixing string | null makes that workaround unnecessary)