feat: implement advanced cross-compilation support and backend customization

[LunaStev]

This PR transforms the Wave compiler into a truly portable toolchain by introducing a robust cross-compilation framework and fine-grained backend control. Key features include expanded architecture support for ARM64 and x86_64, a comprehensive CLI for LLVM and linker customization, and a modernized preprocessor for platform-specific code.

---

Key Changes

1. Advanced Cross-Compilation Framework

• Parameterized Backend: Introduced BackendOptions to manage the target triple, CPU, features, ABI, sysroot, and linker settings.
• Global Target Initialization: Updated the backend to initialize all supported LLVM targets, enabling any host to compile for any supported target.
• ABI support: Refactored abi_c.rs to support the standard System V ABI for both x86_64 and ARM64 across Linux and macOS.

2. Architecture-Specific Inline Assembly

• ARM64 Mapping: Implemented full register group mapping for ARM64 (e.g., x0 through x30, sp) in the assembly planner.
• Dialect Support: The compiler now automatically selects the appropriate assembly dialect (Intel for x86, AT&T/Standard for ARM) based on the target architecture.
• Clobber Optimization: Refined the automatic clobber logic to avoid trashing general-purpose registers during empty or barrier-style assembly blocks.

3. Granular CLI Interface

• LLVM Control: Added the --llvm subcommand to expose low-level backend options: --target, --cpu, --features, --abi, and --sysroot.
• Linker Customization: Introduced -C flags for advanced linking: linker=<path>, link-arg=<arg>, and no-default-libs.
• Extensibility: Reserved the --whale flag for future alternative backend integrations.

4. Robust #[target] Preprocessing

• Awareness Scanner: Significantly improved the #[target] preprocessor. It is now fully aware of block comments, string/char literals, and nested braces, ensuring that platform-specific code blocks are parsed accurately without premature truncation.

5. Infrastructure & Policy

• Tiered Platform Policy: Formally established a Tier 1–4 support matrix in README.md to set clear expectations for target stability.
• Build Orchestration: Updated x.py to include aarch64-unknown-linux-gnu and improved build artifact hygiene by streamlining the object file placement during the linking phase.

---

Example Usage

Cross-compiling for ARM64 Linux from an x86 host:

wavec build main.wave \
    --llvm-target aarch64-unknown-linux-gnu \
    -C linker=aarch64-linux-gnu-gcc

Using the improved preprocessor:

#[target(os="linux")]
fun init() {
    /* Nested braces { and strings "}" no longer break the preprocessor */
    println("Initializing Linux...");
}

---

Benefits

• Portability: Allows developers to target embedded systems, ARM-based servers, and Apple Silicon hardware from a single development environment.
• Precision: Fine-grained CLI control over CPU features and sysroots enables the creation of highly optimized and specialized binaries.
• Robustness: The enhanced preprocessor eliminates subtle bugs when managing large, platform-specific codebases.