code : normalize enum names

[ggerganov]

I noticed we haven't been very consistent with naming the enums. For example:

    // here we repeat the full enum name and add suffix
    enum llama_token_type {
        LLAMA_TOKEN_TYPE_UNDEFINED    = 0,
        LLAMA_TOKEN_TYPE_NORMAL       = 1,
        LLAMA_TOKEN_TYPE_UNKNOWN      = 2,
        LLAMA_TOKEN_TYPE_CONTROL      = 3,
        LLAMA_TOKEN_TYPE_USER_DEFINED = 4,
        LLAMA_TOKEN_TYPE_UNUSED       = 5,
        LLAMA_TOKEN_TYPE_BYTE         = 6,
    };

    // in contrast, here we omit "_type" from the enum name before adding the suffix
    enum ggml_task_type {
        GGML_TASK_INIT = 0,
        GGML_TASK_COMPUTE,
        GGML_TASK_FINALIZE,
    };

With this PR, I propose that we always follow the convention to suffix the full enum name:

enum some_enum_name {
    SOME_ENUM_NAME_VAL_0, // OK
    SOME_ENUM_VAL_0,      // not OK
};

Is this change worth it? It touches a lot of lines

[slaren]

Personally I would prefer to do it the other way around, and remove all the _TYPE _NAME _MODE etc suffixes from the enum values, but ultimately I think that the most important is to be consistent with the style.