Add verbose flag for model loading

[grahameth]

Right now, the console is spammed with a lot of information about the model when llama_load_model_from_file is called. This PR adds a verbose flag so that you can optionally disable this.

[ggerganov]

Isn't redirecting stderr to /dev/null good enough?

[grahameth]

Well, if you just use the provided examples, sure. But in my usecase, I use llama.cpp in a bigger application, and I don't want to disable the whole error stream just for this function call.
But I understand if you don't want this added complexity if you have no use for it, then I'll just change it in my fork.

[bullno1]

I think a better solution would be to be able to specify a log handler with a similar signature: (void* ctx, const char* fmt, ...)

[grahameth]

I can implement it in such a way. Should I do it? I don't wanna do it though if it won't get merged anyway

[ggerganov]

@grahameth I can't promise you that it will be merged, but it is more likely to get merged than the verbose flag

[grahameth]

Alright, I made a new PR for it: #2234