fix: make embedding work again

[vansangpfiev]

Changes in llama.cpp require our interface changes.
To load the embedding model, need to set embedding flag to true, for example:

{
    "llama_model_path": "e:/workspace/model/nomic-embed-text-v1.5.f16.gguf",
    "ctx_len": 5,
    "ngl": 0,
    "n_parallel": 1,
    "embedding": true
} 

[tikikun]

LGTM