port: clef decision-model support + llama-clef-demo (ABI-compatible with b10453.4)

[Siddhesh2377]

What

Ports the clef decision-model architecture onto runanywhere-b10453.4 and wires the joint decision head end to end, plus a CLI that exercises it.

A clef GGUF (general.architecture = clef) attaches a joint decision head: one forward pass over a prompt carrying a state + every typed question + every option scores all options at once. The head reads per-token span tags to pool the right runs, and the per-option scores come back as rows of the embeddings output.

Commits

1. clef model-side port — LLM_ARCH_CLEF registration, arch/KV/tensor tables, llama_model_clef (Qwen3.5 backbone + joint head), rope/memory/graph routing, model saver opt-out, arch test, conversion/clef.py + gguf-py metadata.
2. compat fixes for this base — arch-table duplicate removal, span array inits, decision-order enum in the public header, the missing build_qkv overload the head calls.
3. llama_batch_set_decision_order — span array stays NULL after llama_batch_init and is allocated on first setter call. NULL means "no spans", which is what keeps plain generation on a decision model on the 0.0f status path instead of the all-zeros NaN path.
4. llama-clef-demo — a CLI that runs a real joint decision: spans tagged, one llama_encode, one score per option, softmax → choice/confidence. Named demos + generic --question/--option.

Verification

Built with -DLLAMA_BUILD_TESTS=ON and run against Clef-Flash-Q4_K_M.gguf:

• test-llama-archs lists clef; model loads all 587 tensors with no arch errors.
• llama-clef-demo --demo → A (sqlite-vec) at p=0.92, exit 0.
• Plain generation (no spans) still takes the clean path.

Notes

• Model-side fidelity: src/models/clef.cpp and conversion/clef.py are the same content as the upstream clef support, adapted to this base's batch model.
• The server-side decision endpoints are intentionally not ported; this base has no extended-batch framework and the SDK consumes the head through its own engine op.
• Maple entries untouched; clef is a dense Qwen3.5-9B, no interaction expected.

---

Summary by cubic

Ports the clef decision-model architecture onto runanywhere-b10453.4 so a clef GGUF scores all options in one forward pass. Plain generation on a decision model keeps its previous behavior: batches without tagged spans take the clean path and the head scores nothing.

What's included

• Registers LLM_ARCH_CLEF with a Qwen3.5 backbone plus joint decision head, model mapping, rope/memory/graph routing, arch test, conversion/clef.py, and gguf-py metadata.
• Adds llama_batch_set_decision_order to tag token spans; the span array stays NULL (meaning "no spans") until the first setter call, so batches without spans keep the clean 0.0f status path instead of the invalid-input NaN path.
• Adds llama-clef-demo, a CLI that tags question/option spans, runs one llama_encode, and reads one score per option back from the embeddings output with softmax → choice/confidence, including named demos and generic --question/--option input.

Not included

• Server-side decision endpoints, which need the extended-batch framework this base lacks.
• Model saver support for clef — the head tensors are not saved. Maple entries are untouched.

^{Written for commit adcfbd7. Summary will update on new commits.}

[Siddhesh2377]

Retargeting: the branch now lives in this repo (RunanywhereAI/llama.cpp:clef-decision-port). Opening the in-repo PR next.