Support nullable UTF-8 string arguments in FFI signatures #1033

Summary

Add an FFI descriptor that accepts either ECMAScript text or null, marshalling strings as temporary NUL-terminated UTF-8 and null as a native null pointer.

Why

Many C APIs use nullable const char * parameters. GocciaScript currently forces a binding to choose between utf8string (convenient text, no null-pointer semantics) and pointer (null-capable, rejects JavaScript strings). GocciaScript-Raylib therefore emits seven additional Raw bindings for nullable C-string parameters; see the merged binding documentation.

Current behavior

On current main (34cc32f), using the existing FFI fixture:

const lib = FFI.open("./fixtures/ffi/libfixture" + FFI.suffix);
const strlen = lib.bind("string_length", {
  args: ["utf8string"],
  returns: "i32",
});
console.log(strlen(null));

This prints 4: null is coerced to the text "null" and native code receives that string, not NULL.

Current-source evidence:

• utf8string marshals through AValue.ToStringLiteral.
• JavaScript null stringifies to "null".
• The pointer path accepts null but not JavaScript strings.
• Type parsing accepts one scalar/type descriptor, with no nullable composition.

The fresh-main FFI suite passes in interpreted and bytecode modes (92 tests, 337 assertions); a direct fresh-main stdin reproduction produced 4.

Expected behavior

A nullable UTF-8 string descriptor should marshal strings with the existing strict UTF-8/NUL validation and marshal null as a null pointer. Plain utf8string behavior may remain unchanged for compatibility if the nullable contract is explicit.

Scope notes

• Preserve the ownership distinction: this is for borrowed nullable const char * inputs, not mutable or owned char * buffers.
• Do not make arbitrary pointer descriptors coerce strings implicitly.
• Related: Add explicit FFI support for native shared libraries #178, FFI v2 #827, and GocciaScript-Raylib PR #7.