native-wasmtime: Rust/wasmtime + wgpu reference backend

[Milky2018]

This adds a second native runtime backend implemented in Rust (Cargo), intended as a side-by-side reference implementation alongside the existing MoonBit/wasmoon runtime.

Changes:

• New crate: mgstudio-runtime/native-wasmtime (wasmtime + winit + wgpu)
• mgstudio-cli: add --backend wasmtime (default remains wasmoon)
• mgstudio-cli: auto-build/run the Rust runtime (with overrides via MGSTUDIO_WASMTIME_RUNTIME and MGSTUDIO_WASMTIME_CARGO_TARGET)
• Docs: usage + Apple Silicon notes

Notes:

• On macOS arm64 (Apple Silicon), some wasm-gc carts may currently trigger a Cranelift AArch64 panic during compilation. A practical workaround is building the runtime for x86_64-apple-darwin and running under Rosetta.

[Milky2018]

Thanks for putting this together — having a reference Wasmtime/WGPU backend is useful for comparison.

However, I don’t think this PR is in a mergeable state for the main branch right now, mainly because it’s out of sync with the current direction of the repo:

1. Outdated vs current SDK + moon.game.json design

• Current main uses moon.game.json.sdkroot + CLI-built assets overlay and stages mgstudio-runtime-web.js from the SDK.
• This PR’s mgstudio-cli/main/cmd_run.mbt relies on cfg.native_wgpu_lib (no longer exists) and introduces auto-discovery/build logic for the Rust runtime (searching upward for Cargo.toml). That conflicts with the “single source of truth” config approach.

2. Introduces a second native runtime implemented in Rust (new toolchain + separate exe)

• The repo currently focuses on a MoonBit native runtime (wasmoon JIT) and an SDK packaging story for darwin-arm64.
• Adding a Rust runtime (plus Cargo.lock and a large amount of Rust code) significantly increases maintenance/CI scope. If we want this long-term, it probably belongs in a separate repo or behind an explicit experimental gate.

3. Rebase required

• As a single commit against an older base, it would effectively revert/remove multiple recent changes on main (SDK scripts/workflows, credits/licensing updates, etc.) when merged.

If you’d like to continue with this backend, I’d suggest:

• Rebase onto the latest main first.
• Remove repo-root auto-discovery/build behavior (no implicit upward search). The CLI should use moon.game.json.sdkroot only.
• Decide whether the Rust runtime is (a) a separate optional component/repo, or (b) an experimental backend not shipped in the default SDK.
• Add an explicit licensing/credits plan for the Rust dependency tree (wasmtime/wgpu/winit/etc.) if we plan to distribute binaries.

Given the above, my recommendation is: do not merge as-is; either convert to draft and rework, or move this backend to a separate experimental repository.

[Milky2018]

Addressed the concerns from the review:

• Rebased onto latest main.
• Updated mgstudio-cli/main/cmd_run.mbt to match the current SDK + assets overlay flow (uses resolve_sdk_layout + staged overlay assets).
• Removed repo-root auto-discovery and implicit cargo build behavior for the wasmtime backend.
  • CLI now looks for <sdkroot>/bin/mgstudio-runtime-native-wasmtime (optional/experimental), or you can explicitly set MGSTUDIO_WASMTIME_RUNTIME=/abs/path/to/mgstudio-runtime-native-wasmtime.
• Updated CLI docs + runtime README accordingly, and added a concrete note/plan for third-party notices if we ever ship this runtime in an SDK (mgstudio-runtime/native-wasmtime/THIRD_PARTY_NOTICES.md).