feat: introduce a ComputeBackend seam to abstract the forward-execution engine

[inureyes]

Summary

Introduce a ComputeBackend seam so a future non-MLX execution engine can host forward() without routing through the MLX bridge. The motivating target is FuriosaAI TCP / RNGD, whose furiosa-opt toolchain compiles to a virtual ISA and cannot use MLX at all. This PR lands the seam only and is backend-neutral; it implements no Furiosa or TCP kernels.

Seam design

ComputeBackend (new src/backend/mod.rs) abstracts who executes forward(), not individual ops. It sits at the model-load boundary (called once per load), not at the per-token forward() call, so MLX graph fusion and mx.compile stay intact and no indirection enters the inner loop. The trait is drawn narrowly to the load entry points (load_model, load_model_with_adapter, load_model_with_tensor_parallel) so the MLX path keeps its concrete hot types exposed: paged KV, prompt-cache detach and adopt, and cache tensors are never type-erased behind Box<dyn>. A non-MLX engine implements the same LanguageModel forward contract behind a different backend.

What moved behind MlxBackend

MlxBackend (src/backend/mlx.rs) is a zero-sized type that implements ComputeBackend by delegating unchanged to the existing crate::loading entry points. No loading logic is reimplemented; the same code runs whether reached directly or through the seam. The public load_model / load_model_with_adapter / load_model_with_tensor_parallel functions remain available (tests and bench harnesses still use them).

Feature flag

A default-off experimental-backend Cargo feature gates the optional non-MLX path. The src/backend/experimental.rs scaffold module and the Backend::Experimental enum variant are cfg-gated behind it, so shipping binaries (Apple Silicon, CUDA) compile no extra backend code. The scaffold reports "not implemented" rather than pretending to load; a real engine (and any hardware-feasibility gate) is future work. The feature-on select_backend() is the only place that reads a runtime backend switch (MLXCEL_BACKEND), and it is compiled only when the feature is enabled.

How codegen equivalence / no dispatch is guaranteed when the feature is off

Backend is an enum whose only variant under default features is Backend::Mlx, and MlxBackend is zero-sized, so Backend is itself zero-sized with no discriminant. select_backend() (#[inline], #[must_use]) is a constant constructor that always returns that one variant with no environment read and no branch. Every Backend method is a single-arm match marked #[inline]. After inlining (release builds use fat LTO and codegen-units = 1), select_backend().load_model(p) lowers to a direct call to the existing MLX loader, identical to the pre-seam build. The env-reading selection path and the second enum variant only exist under #[cfg(feature = "experimental-backend")], so they are absent from default codegen entirely.

How behavior preservation is ensured

Behavior is preserved by construction: the control-plane load call sites now go through the seam, and the seam methods delegate to the unchanged loaders, so the same loading and forward code runs. Rerouted call sites: src/commands/generate.rs (primary load: tensor-parallel / adapter / plain, plus the offline draft-model load), src/commands/chat.rs (REPL load), and src/server/model_worker.rs (both the batched and the legacy sequential worker loops, each covering tensor-parallel / adapter / plain). The pipeline-parallel branch keeps its own distributed loader and does not go through the seam.

What changed

• src/backend/mod.rs (new): ComputeBackend trait, Backend enum, select_backend().
• src/backend/mlx.rs (new): MlxBackend, delegates to crate::loading.
• src/backend/experimental.rs (new, cfg-gated): scaffold plug-in slot for a non-MLX engine.
• src/backend/tests.rs (new): scoped seam tests.
• src/lib.rs: pub mod backend; and re-export of Backend, ComputeBackend, MlxBackend, select_backend.
• Cargo.toml: default-off experimental-backend feature.
• src/commands/generate.rs, src/commands/chat.rs, src/server/model_worker.rs: route loads through select_backend().

Test plan

• cargo check --lib --features metal,accelerate
• cargo check --bins --features metal,accelerate
• cargo check --lib --features metal,accelerate,experimental-backend (gated module compiles)
• cargo clippy --lib --tests --features metal,accelerate -- -D warnings
• cargo clippy --lib --features metal,accelerate,experimental-backend -- -D warnings
• cargo test --lib backend:: --features metal,accelerate (3 tests pass)
• cargo fmt --check
• Real-checkpoint temp-0 byte-identical token parity and throughput on the MLX path are owned by the maintainer's release-build parity gate.

Closes #338