llama_context_params do not match latest llama.cpp code

[sandro-qiang]

The latest definition of llama_context_params, added flash_attn_type and removed flash_attn, and added kv_unified.

struct llama_context_params {
        uint32_t n_ctx;             // text context, 0 = from model
        uint32_t n_batch;           // logical maximum batch size that can be submitted to llama_decode
        uint32_t n_ubatch;          // physical maximum batch size
        uint32_t n_seq_max;         // max number of sequences (i.e. distinct states for recurrent models)
        int32_t  n_threads;         // number of threads to use for generation
        int32_t  n_threads_batch;   // number of threads to use for batch processing

        enum llama_rope_scaling_type rope_scaling_type; // RoPE scaling type, from `enum llama_rope_scaling_type`
        enum llama_pooling_type      pooling_type;      // whether to pool (sum) embedding results by sequence id
        enum llama_attention_type    attention_type;    // attention type to use for embeddings
        enum llama_flash_attn_type   flash_attn_type;   // when to enable Flash Attention

        // ref: https://github.com/ggml-org/llama.cpp/pull/2054
        float    rope_freq_base;   // RoPE base frequency, 0 = from model
        float    rope_freq_scale;  // RoPE frequency scaling factor, 0 = from model
        float    yarn_ext_factor;  // YaRN extrapolation mix factor, negative = from model
        float    yarn_attn_factor; // YaRN magnitude scaling factor
        float    yarn_beta_fast;   // YaRN low correction dim
        float    yarn_beta_slow;   // YaRN high correction dim
        uint32_t yarn_orig_ctx;    // YaRN original context size
        float    defrag_thold;     // [DEPRECATED] defragment the KV cache if holes/size > thold, <= 0 disabled (default)

        ggml_backend_sched_eval_callback cb_eval;
        void * cb_eval_user_data;

        enum ggml_type type_k; // data type for K cache [EXPERIMENTAL]
        enum ggml_type type_v; // data type for V cache [EXPERIMENTAL]

        // Abort callback
        // if it returns true, execution of llama_decode() will be aborted
        // currently works only with CPU execution
        ggml_abort_callback abort_callback;
        void *              abort_callback_data;

        // Keep the booleans together and at the end of the struct to avoid misalignment during copy-by-value.
        bool embeddings;  // if true, extract embeddings (together with logits)
        bool offload_kqv; // offload the KQV ops (including the KV cache) to GPU
        bool no_perf;     // measure performance timings
        bool op_offload;  // offload host tensor operations to device
        bool swa_full;    // use full-size SWA cache (https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055)
                          // NOTE: setting to false when n_seq_max > 1 can cause bad performance in some cases
                          //       ref: https://github.com/ggml-org/llama.cpp/pull/13845#issuecomment-2924800573
        bool kv_unified;  // use a unified buffer across the input sequences when computing the attention
                          // try to disable when n_seq_max > 1 for improved performance when the sequences do not share a large prefix
                          // ref: https://github.com/ggml-org/llama.cpp/pull/14363
    };

The wrong definiton make program core dump immediately when loading model.

May be easy_llama can tag the same as llama.cpp, e.g. b6324. This can make user download corresponding llama.cpp code easier.