Allow native aggregate fields named buffer and byteOffset #1032

Summary

Allow FFI structure and union descriptors to preserve native field names buffer and byteOffset without colliding with aggregate-value metadata.

Why

These are ordinary C field names. Raylib 6.0's AudioStream.buffer must currently be renamed to nativeBuffer in generated JavaScript, with a separate alias table to recover the native name. See the merged binding documentation.

Current behavior

On current main (34cc32f):

FFI.struct({ buffer: "pointer" });

Creation throws:

FFI aggregate field name is reserved: buffer

byteOffset is rejected the same way.

Current-source evidence:

• Aggregate construction rejects both names.
• The focused tests explicitly require both rejections.
• The names are currently used for aggregate backing-view metadata.

The fresh-main FFI suite passes in interpreted and bytecode modes (92 tests, 337 assertions), confirming the collision is intentional current behavior.

Expected behavior

Generated FFI aggregates should expose the exact native fields buffer and byteOffset while retaining an unambiguous way to access backing-buffer metadata. The solution should preserve nested-view and detached-buffer safety.

Scope notes

• Do not rename native fields in generated bindings as the permanent API.
• This is an FFI aggregate object-model collision, not a JavaScript reserved-keyword issue.
• Related: FFI: Support struct type declarations and marshalling #215, FFI: Support union type declarations #216, FFI v2 #827, and GocciaScript-Raylib PR #7.