Feature Request: Make llm_arch_supports_mixed_batch suitable for C and FFI consumers #30026

Prerequisites

• I am running the latest code. Mention the version if possible as well.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new and useful enhancement to share.

Feature Description

I need access to this API. Here are some details:

• PR llama: support both embd + raw tokens in batch #29622 added support for batches containing both raw token entries and embedding entries.
• Some model architectures do not support these mixed batches.
• llama.cpp checks this internally with llm_arch_supports_mixed_batch.
• llm_arch_supports_mixed_batch is not part of the public C API and takes the private llm_arch type, so Go/libffi consumers cannot safely bind it.
• Request a supported C API that reports whether a loaded model supports mixed token-and-embedding batches.

Here is some related work and sources

• Mixed-batch implementation:
  llama: support both embd + raw tokens in batch #29622
• Merge commit:
  0bb496d
• Internal helper declaration:
  https://github.com/ggml-org/llama.cpp/blob/master/src/llama-arch.h
• Internal helper implementation:
  https://github.com/ggml-org/llama.cpp/blob/master/src/llama-arch.cpp
• Public extended batch API:
  https://github.com/ggml-org/llama.cpp/blob/master/include/llama.h#L1007-L1090
• Yzma extended batch binding:
  https://github.com/hybridgroup/yzma/blob/main/pkg/llama/batch_ext.go

Motivation

Here are my motivations

• Yzma binds llama_batch_ext and llama_process through Go/libffi.
• Yzma needs to know whether a loaded model supports a mixed batch before submitting one.
• The internal C++ helper cannot be safely bound because it is not public C API and uses the private llm_arch type.
• Copying llama.cpp's architecture exception list into Yzma would become stale as architectures are added or changed.
• A public model capability query would let Yzma expose the same decision safely to Go callers

Possible Implementation

Here are some ideas for a possible implementation

A small public C wrapper around the existing internal check, following the existing model capability APIs:

LLAMA_API bool llama_model_supports_mixed_batch(
        const struct llama_model * model);

The implementation can call llm_arch_supports_mixed_batch(model->arch) internally. The exact API name belongs to llama.cpp maintainers.