A Frontend for Safe GPU Offloading in Rust

The `offload` feature is currently working on the Rust compiler but still there's no safe and user-friendly interface. Right now, the only way to use it is via the `offload` intrinsic, which is discouraged for general use. The goal of this project is to build an ergonomic frontend around a function-like macro together with lightweight abstractions for device management, host-to-device data handling through Device<T>, and compile-time validation of device-compatible types.