Deprecated shim over llama.cpp model listing.

The GET /v1/models transport now lives in ExAthena.ModelListing and is reached as ExAthena.list_models(:llamacpp), so that every backend is enumerated through one code path instead of one helper module per server.

Retained for existing callers, returning the same sorted strings and error atoms as before. The llama.cpp server (llama-server) typically serves one model at a time, so the list usually has a single entry.